principles of multi-slice cardiac ct imaging 4

71

Principles of Multi-slice Cardiac CT Imaging 4Bernd Ohnesorge and Thomas Flohr

C o n t e n t s

4.1 Basic Performance Requirements for CT Imaging of the Heart 71

4.2 CT Imaging with Optimized Temporal Resolution: The Principle of Half-Scan Reconstruction 73

4.3 Prospectively ECG-Triggered Multi-slice CT 76

4.4 Retrospectively ECG-Gated Multi-slice CT 794.4.1 Multi-slice Cardiac Spiral Reconstruction 814.4.2 ECG-Gated Multi-slice Spiral Acquisition 834.4.3 Segmented Cardiac Reconstruction Algorithms 884.4.4 Cardiac Cone-Beam Reconstruction Algorithms 974.4.5 ECG-Gated Spiral Scanning with Increased Volume Speed 101

4.5 Synchronization with the ECG and Cardiac Motion 1064.5.1 ECG-Based Phase Selection 1064.5.2 The Pros and Cons of ECG Gating and ECG Triggering 1084.5.3 Alternative Cardiac-Motion Gating Approaches 110

4.6 Radiation Exposure Considerations 1134.6.1 Principles of Radiation Dose Measurement in CT 1134.6.2 Radiation Exposure for Selected Cardiac Examination Protocols 1174.6.3 Exposure Reduction with ECG-Gated Tube-Current Modulation 1194.6.4 Optimization for Different Patient Sizes 1214.6.5 Optimization of Contrast-to-Noise Ratio 123

References 124

4.1 Basic Performance Requirements for CT Imaging of the Heart

To virtually freeze cardiac motion and avoid motion-artifacts, very short exposure times are needed for the acquisition of transaxial slices. High temporal resolution is particularly important for imaging the coronary arteries, as they are located very close to the heart muscles and show strong movement during the cardiac cycle. Due to the very complex 3D motion pattern of the heart, the intensity of move-ment varies for different cardiac anatomies and dif-ferent coronary vessels, and within the cardiac cycle. The strongest movement is present during contrac-tion of the atria and ventricles in systole. The short end-systolic rest phase is followed by a continuous-filling phase of the ventricles during diastole that slows down towards mid- and end-diastole. This motion pattern can be measured by the dislocation of the left ventricular wall, the aortic valve flaps, and the different segments of the coronary arteries in representative transaxial planes (Fig. 4.1a) (Achen-bach 2000). The least amount of movement of the major cardiac anatomy and the coronary arteries, and thus the least amount of dislocation over time, is observed in end-systole and mid- to end-diastole of the cardiac cycle (Fig. 4.1a). While the duration of the end-systolic phase (approximately 100–150 ms) is more or less independent of the present heart rate (and the related RR-interval time), the duration of the phase with the least cardiac motion during dias-tole narrows with increasing heart rate (Fig. 4.1b). For low heart rates, the end-systolic phase is much shorter than the diastolic rest phase, but the dif-ference decreases with increasing heart rate. For

72 Chapter 4 · Principles of Multi-slice Cardiac CT Imaging

higher heart rates, the diastolic rest phase can be even shorter than the end-systolic rest phase.

The cardiac anatomy should be imaged during those phases of cardiac cycle in which there is the least movement; that is, when the least blurring due to motion artifacts is to be expected. Thus, image acquisition and reconstruction need to be synchro-nized as accurately as possible to the movement of the heart, i.e., by using ECG information that is recorded in parallel to the CT scan acquisition. Since the fastest gantry rotation speeds of today’s multi-slice CT scanners are between 0.5 s up to 0.33 s per rotation, imaging of the heart is usually performed during the diastolic rest phase of the cardiac cycle, as the achievable temporal resolution is usually not sufficient to reliably eliminate cardiac motion during other phases of the cardiac cycle. However, with CT gantry rotation speeds becoming even faster, imaging in the end-systolic rest phase may also become feasible.

According to a rough estimation, a temporal res-olution of about 250 ms is appropriate for motion-free imaging in the diastolic rest phase up to a heart rate of about 60 bpm, about 200 ms up to a heart rate of 70 bpm, and approximately 150 ms for clinically

usual heart rates up to 90 bpm (Fig. 4.1b). It can be expected that a temporal resolution of about 100 ms is sufficient for imaging the heart during the dia-stolic or end-systolic rest phase also at high heart rates (Achenbach 2000). Image acquisition during other phases of the cardiac cycle (e.g., systole) with rapid cardiac motion is needed for evaluation of car-diac function. Motion-free imaging during phases other than the diastolic and end-systolic rest phases requires a temporal resolution of about 50 ms (Stehling 1991) and is usually not possible with state-of-the-art CT systems. Thus, only the larger cardiac anatomy or cardiac anatomy with low motion amplitudes should be assessed during phases of high cardiac motion.

Many structures of the cardiac morphology, especially the coronary arteries and the cardiac valves including the valve flaps, represent small and complex 3D structures that require very high and, at best, sub-millimeter isotropic spatial reso-lution with longitudinal resolution close or equal to in-plane resolution. The most proximal coro-nary segments and the distal segments of the right coronary artery (RCA) are directed parallel to the image plane, while the middle segments are directed

wall

valce

LADRCA

2

0

-2

-4

-6

-8

-10

-12

-14

-16

0 20 40 60 80 100RR-Phase / % RR-Time

Def

lect

ion

/ m

m

R R

Q QS S

T P

bpm

ms sm052sm052

Diastole

500

400

300

200

100

040 50 60 70 80 90 100 110 120

Exposure Time

250ms200ms

150ms100ms

Fig. 4.1. a Cardiac motion during the different phases of the cardiac cycle (Achenbach 2000). Rest phases of the myocar-dial wall and the coronary arteries can be observed during diastolic fi lling and during end-systole. b The duration of the phase of lowest cardiac motion, during the diastolic rest phase, depends on heart rate. For clinically usual heart rates up to 90 bpm, a temporal resolution of about 150 ms is needed for motion-free imaging during the diastolic rest phase, whereas a 100-ms temporal resolution is feasible for imaging during the cardiac rest phases at heart rates above 100 bpma

b

73

perpendicular to the image plane. The lumen diam-eter of the main segments of the coronary artery tree ranges from 4 mm (left main coronary artery) down to about 1 mm (peripheral left main coronary artery and circumflex). Detection and quantifica-tion of coronary stenosis with the ability to differ-entiate a 20% change in the diameter of the vessel lumen for the larger-caliber vessels represents a viable goal for cardiac CT imaging. To achieve this, CT systems need to provide a spatial resolution in all three dimensions (isotropic) of at least 0.5 mm for visualization of the main coronary vessels and of smaller branches. Thus, in-plane and through-plane resolution significantly below 1 mm are needed to assess the main coronary segments, including nar-rowing and plaques. Less spatial resolution is suf-ficient for assessment of the larger cardiac anatomy, such as the myocardium and the cardiac chambers. However, scans should a priori be acquired at best possible spatial resolution, as lower spatial resolu-tion can be generated via data post-processing.

In addition to high spatial resolution, sufficient contrast-to-noise ratio is important to resolve small and low-contrast structures, such as atherosclerotic coronary plaques, with different attenuation prop-erties. Appropriate low-contrast resolution has to be provided with limited radiation exposure at the shortest possible exposure time. For most cardiac applications, appropriate contrast enhancement of the cardiac vessels and the cardiac anatomy is achieved with peripheral injection of contrast agent and optimized timing of the bolus. Sufficient con-trast enhancement is particularly important for imaging the distal coronary segments, as they are located very close to the myocardium and are sepa-rated from it only by a thin layer of epicardial fat. Only the assessment of cardiac and coronary calcifi-cation is possible without administration of contrast agents.

Scan acquisition within a single and short breath-hold time is mandatory for minimizing the amount of contrast material needed for vascular enhance-ment and to avoid respiratory artifacts. Breath-hold times of 20 s or less are appropriate for stable patients but 10 s or less are preferable for less stable, dyspneic patients. Ideally, a complete data set of the entire heart anatomy would be acquired within a single phase of the cardiac cycle without patient move-

ment, but CT technology today, and presumably in the near future, does not provide sufficient detec-tor width to cover the entire heart volume within a single heartbeat. Instead, images at consecutive z-positions that continuously cover the heart volume need to be generated from data acquired in different cardiac cycles. A virtually «frozen» cardiac volume can only be produced by phase-consistent synchro-nization of the acquisition to the movement of the heart using simultaneously recorded ECG informa-tion. A reasonably stable sinus rhythm without rapid arrhythmic changes provides the best 3D images of the cardiac anatomy. Functional information can be derived if cardiac volume images can be gener-ated in different phases of the cardiac cycle, e.g., as a basis for quantitative evaluation.

Cardiac imaging is obviously a highly demand-ing application for CT, since the temporal, spatial, contrast resolutions as well as scan time have to be optimized simultaneously, and radiation exposure has to be limited to levels of related imaging modali-ties, such as invasive diagnostic coronary angiogra-phy or nuclear imaging studies. These conflicting performance requirements have to be fulfilled and optimized for the particular application at the same time and within the same cardiac scan protocol. Optimization of one performance cornerstone alone (e.g., high temporal resolution, low radiation expo-sure) by trading-off other important parameters (e.g., spatial and contrast resolutions) may not lead to clinically useful results.

4.2 CT Imaging with Optimized Temporal Resolution: The Principle of Half-Scan Reconstruction

For motion-free imaging of the heart, data have to be acquired during phases of the cardiac cycle with little cardiac motion and with as high as possible temporal resolution. In the first place, temporal resolution of an axial CT slice is determined by the exposure time associated with the scan data used for reconstruction of that CT slice. Therefore, scan techniques and reconstruction algorithms need

4.2 · CT Imaging with Optimized Temporal Resolution: The Principle of Half-Scan Reconstruction


to be developed that use a minimum amount of scan data while maintaining high image quality. In modern 3rd-generation CT systems with fan-beam geometry, the minimum amount of scan data needed for reconstruction of axial slices is a so-called partial scan (Parker 1982). Depending on the exact system geometry and its dimensions, a partial-scan fan-beam data set has to cover a pro-jection-angle interval αP (angle interval between tube positions at the start and end points of tube rotation) of 180° plus the breadth of the X-ray fan: αP = π + βf. The breadth of the X-ray fan-beam βf depends strongly on the diameter of the scan field of view (usually 50 cm) and the distances of the focal spot and detector from the center of the scan field of view. The equation αP = π + βf states that a minimum data segment of 180° has to be avail-

able for every fan angle β. As a consequence, a partial rotation usually covers about two-thirds of a rotation (≈240°) (Fig. 4.2a). Conventional partial-scan reconstruction techniques based on fan-beam geometry (Parker 1982) make use of all acquired data even if more data than the minimum angle of 180° are available for a fan angle β. These tech-niques produce a temporal resolution equal to the partial-scan acquisition time (≈two-thirds of the rotation time).

Better temporal resolution can be achieved with special reconstruction algorithms that use the min-imum required amount of scan data. These algo-rithms, referred to as "half-scan" reconstruction, can be best explained using parallel-beam geom-etry. To this end, the fan-beam geometry of the par-tial-scan data set is transformed to parallel-beam

α

β

β

Ρ

Θ

Rebinning

Rebinning

Ρ

π

Θα

π

2π 2π

Transition Ρƒ(π+βƒ…π+βƒ+αΤ,β)

Transition ΡΡ(π…π+ΘΤ,Ρ)

Transition ΡΡ(0…ΘΤ,Ρ)

not used

not used

Fan Beam Geometry Data Parallel Geometry Data

Fig. 4.2a, b. Principle of “half-scan” image reconstruction based on parallel-beam scan data. Scan data acquired in fan-beam geometry (a) with fan-beam projection angle α and fan-ray angle β is transformed into parallel-beam geometry (b) with parallel-beam projection angle Θ and parallel-ray position p using the “rebinning” technique and 2D interpolations. Acquired fan-beam data that form incomplete parallel projections are omitted during reconstruction. A data transition range enables smooth transition weighing for reduction of artifacts due to data inconsistencies

a b

75

geometry using "rebinning" techniques (Kak 1988). The rebinning of a partial-scan fan-beam data set provides 180° of complete parallel projections, including chunks of incomplete parallel projections that consist of redundant data. Fan-beam data are described by angle α of the tube position and angle β of a certain ray within the fan-beam projection (Fig. 4.2a). Correspondingly, parallel data can be identified with angle Θ for the projection angle and position p of the rays within the parallel-beam pro-jection (Fig. 4.2b). Fan-beam data Pf(α,β) are trans-formed into parallel-beam data Pp(Θ,p) based on:

Pp(Θ,p) = Pf(α+β−π/2,RF sinβ) for Θ = 0…π,p = -D/2…D/2 (4.1)

RF represents the distance of the X-ray source to the center of rotation of the measurement system, and D the diameter of the scan field of view. As all projection data are digitally processed and there-fore digitally sampled, continuous transformation requires simple interpolation procedures. The geo-metrical transformation is illustrated in Figure 4.2. An image can be reconstructed from parallel projec-tions that cover a parallel-projection angle range of 180°. If the extra data sectors with incomplete par-allel projections are omitted during reconstruction (Fig. 4.2b), temporal resolution is improved.

However, half-scan reconstruction is prone to streak-type artifacts due to inconsistencies of the projections Pp(0,p) and Pp(π,p) at the beginning and end of the data set that may be produced by object motion between the acquisition of these projections. This effect can be substantially reduced by acquisi-tion of a slightly extended range of fan-beam pro-jections αP = π + βf + αT. The additional projection range αT transforms into an extra range of paral-lel-beam projections ΘT = αT for smooth transition weighting of the projections at the start and end of the parallel-beam data set:

P’p(Θ,p) = W(Θ) Pp(Θ,p) + (1- W(Θ)) Pp(Θ+π,p’) forΘ = 0…ΘT, p = -D/2…D/2 (4.2)

P’p(Θ,p) = Pp(Θ,p)forΘ = ΘT…π, p = -D/2…D/2

p’ = -p represents the so-called complementary ray.

The function W(Θ) [e.g. W(Θ) = sin2(πΘ/2ΘT)] represents a smooth, monotone transition between 0 and 1 within the angle range 0...ΘT. We used a transition range of ΘT = 28° with a total scan angle of αP = 260°. This type of transition weighting does not affect temporal resolution but can effectively suppress streak-artifacts in a reconstruction of the parallel-beam data set P’p(Θ,p).

The temporal resolution that is present at a par-ticular position within the scan field of view is determined by the temporal window of the data that contributes to the reconstruction of that particular image point. The temporal resolution is represented by time sensitivity profiles (Hu 2000) that vary with position in the scan field of view (similar to slice sensitivity profiles for spiral reconstruction). We used the full-width at half maximum (FWHM) of the time sensitivity profile as a measure to determine the distribution of the temporal resolution within the scan field of view produced by our reconstruc-tion algorithm. The temporal resolution equals half the rotation time in a centered region of the scan field of view (e.g. 250 ms for 500-ms rotation time, 200 ms for 400-ms rotation time, and 165 ms for 330-ms rotation time). It shows an approximately linear decrease between the edges of the scan field of view, with a gradient that is perpendicular to the connec-tion line of the start and end position of the source (Fig. 4.3a). Different temporal data windows visual-ized by the paths of different image points within the fan-beam data set (Fig. 4.3b) can qualitatively explain the different temporal resolutions within the scan field.

In all modern multi-slice CT scanners, the half-scan reconstruction technique is the method of choice for image reconstruction in cardiac applications. In clinical applications, the heart should be sufficiently centered within the scan field of view in order to maintain a consistent and stable temporal resolu-tion of about half the rotation time. It is not possible to make clinical use of the areas with better tempo-ral resolution within the scan field of view because the start and end positions of the X-ray source vary strongly for individual slices during ECG-synchro-nized acquisition and, as they depend on the patient’s heart rate, can therefore not be fixed.

4.2 · CT Imaging with Optimized Temporal Resolution: The Principle of Half-Scan Reconstruction


4.3 Prospectively ECG-Triggered Multi-slice CT

Prospective ECG triggering is an established method to synchronize sequential CT scanning using the partial-scan technique to the motion of the heart in order to acquire data in a certain phase of the cardiac cycle, preferably in the diastolic phase, when cardiac motion is minimal.

For ECG-triggered sequential imaging, a prospec-tive trigger is derived from the ECG trace to initiate the CT scan with a certain, user-selectable delay time after the R-wave. For best temporal resolution, a partial scan is acquired that is reconstructed with the half-scan algorithm described above. After every scan, the table moves by the width of the acquired scan range in the z-direction towards the next scan

position in order to provide gap-less volume cov-erage. The delay time for scan acquisition after an R-wave is calculated from a given phase parameter (e.g., a percentage of the RR-interval time as delay after an R-wave) for each cardiac cycle and is indi-vidually based on a prospective estimation of the RR-intervals. Usually, the delay is defined such that the scans are acquired during the diastolic phase of the heart.

Prospectively ECG-triggered sequential scanning has already been used with electron-beam (EBCT) (Stanford 1992) and mechanical single-slice CT (Becker 1999). In both techniques, single slices are acquired in consecutive heartbeats within equiva-lent phases of the cardiac cycle. The total scan time is related to the heart rate of the patient and is often too long for volume coverage with thin slices within a single breath-hold. Multi-slice CT scanners can

0.15s–0.27s: based on 0.40s rotation time0.13s–0.22s: based on 0.33s rotation time

π

2π

0.27s0.22s

0.24s0.20s

0.22s0.18s

0.20s0.16s 0.18s

0.15s 0.16s0.14s

0.15s0.13s

Scan End

Scan Start

Focus Path

α

β

β

α

0.15

s/0.

13s

0.27

s/0.

22s

Fig. 4.3a, b. Display of a spatially variable temporal resolution after half-scan reconstruction within the scan fi eld of view (examples given for 400- and 330-ms rotation times). a Temporal resolution increases continuously between the edges of the scan fi eld of view with a gradient that is perpendicular to the line connecting the source start and source end positions. b This situation can be explained by temporal data windows of different widths within the acquired fan-beam data set that contribute to image points at different locations during image reconstruction (b)

b

a

77

acquire multiple slices within one heartbeat. Thus, they offer shorter scan times that can be beneficial for clinical applications.

The first generation of multi-slice CT scanners, introduced in 1998, allowed for simultaneous acqui-sition of up to four adjacent slices per prospective ECG trigger for sequential coverage of the heart volume (Fig. 4.4a). Newer, multi-slice CT scanners acquire up to 16 slices simultaneously per heart beat and provide shorter rotation times and shorter acqui-sition windows per heart cycle (Fig. 4.4b). The latest generation of multi-slice CT scanners, with up to 64 slices per rotation, often do not use the maximum possible amount of slices. The maximum number of simultaneously acquired slices in sequential scan mode of such scanners is usually limited to 24–32 slices, due to increasing cone-beam artifacts that cannot easily be compensated by sequential scan-ning. With an increasing number of slices, more scan volume can be covered with every individual scan. This translates either into a significantly shorter acquisition time for the total scan range or the abil-ity to acquire a given scan range with thinner slices in comfortable breathhold times (Fig. 4.4b).

The concept of coverage of the entire heart with sequential ECG-triggered scans is illustrated in Figure 4.5. The exact scan time for such coverage heart depends on the patient’s individual heart rate, as the RR-interval time determines the time between the individual scans. In this regard, an additional technical effect has to be considered. As the patient table has to move by the distance covered by mul-tiple slices in between the scans, a certain techni-cal delay time (so-called scan-cycle time) has to be taken into consideration before a following scan can be initiated. The minimum scan-cycle time between two consecutive ECG-triggered scans depends on the table-feed between the two scans and on the acquisition time of one scan. Usual scan-cycle times of modern multi-slice CT scanners are in the range of 0.8–1.5 s; thus, one heart beat has to be skipped in between every scan for usual clinical examina-tions at heart rates between 50 and 90 bpm with RR-interval times between 0.7 and 1.2 s. In the example of a 16-slice scanner presented in Figure 4.5, the entire heart is scanned within the time of 15 heart beats and reconstructed from data that are acquired during eight heart beats.

The two different filter techniques that are com-monly used for prospective estimation of the posi-tion of the following R-wave are mean filtering (e.g., 3 previous RR-intervals) and median filtering (e.g., 5 previous RR-intervals). The median filter approach shows increased robustness for patients with mod-erate arrhythmia due to the fact that single extra beats are eliminated.

To diagnose dynamic processes, ECG-triggered acquisition can also be done without table-feed in between the scans. The same volume is then acquired in corresponding phases of consecutive heartbeats. As no table-feed is needed, the scan-cycle time is reduced and scans can usually be acquired within every heart beat for normal heart rates (scan-cycle time 0.5–0.8 s).

Prospectively ECG-triggered multi-slice CT acquisition results in significantly faster volume coverage than obtained with ECG-triggered acqui-sition with single-slice mechanical CT or EBCT. The acquisition of a small volume with each triggered scan reduces the probability of misregistration of lesions that can occur due to significant motion of the heart in the z-direction. With 4-slice CT and a collimated slice-width of 4 × 2.5 mm, the complete heart can be scanned within a comfortable single breath-hold of 15–20 s. With this performance, prospective ECG triggering is feasible for quantifi-cation of coronary calcification (Fig. 4.6a) and CT angiographic imaging of larger cardiac and cardio-thoracic anatomy, such as the myocardium, cardiac valves, thoracic aorta, and proximal segments of coronary bypass grafts (Fig. 4.6b). With modern 16- and 64-slice CT, breath-hold times of ECG-trig-gered acquisition can be further reduced and the use of thin collimation becomes feasible. The increased rotation speed of up to 0.33 s per rotation provides increased temporal resolution up to 165 ms for sig-nificantly improved robustness compared to 4- and 8-slice CT, also at higher heart rates. Furthermore, with 16- and 64-slice CT scanners prospective ECG-triggered scanning can be used for quantification of coronary calcification (Fig. 4.6c) as well as exami-nation of the general cardiac anatomy and coronary bypass grafts (Fig. 4.6d). However, the quality of contrast-enhanced imaging of smaller cardiac anat-omy and the coronary arteries with ECG triggering is limited, since the longitudinal spatial resolution

4.3 · Prospectively ECG-Triggered Multi-slice CT


Fig. 4.4a, b. Sequential volume coverage with prospectively ECG-triggered 4-slice (a) and 16-slice (b) acquisition. Four and 16 adjacent images, respectively, with a slice thickness depending on the collimation (hatched blocks) are acquired at a time with Trot/2 temporal resolution. Due to the limitation of the scan-cycle time, a scan can be acquired every other heart cycle for usual heart rates. Faster volume coverage and thinner slices are provided by 16-slice CT

Time

Sequential Images Trot / 2 Temp. Res.

Delay Delay

R R R R

Scan

Scan

ECG

Table Feed

Time

Delay Delay

R R R R

Scan

Scan

ECG

Table Feed

Sequential Images Trot / 2 Temp. Res.

z - P

osit

ion

z - P

osit

ion

a

b

79

is compromised by the purely sequential, non-over-lapping slice acquisition. In addition, small changes in heart rate during the scan can cause acquisition in inconsistent heart phases and thus inconsistent volume coverage, resulting in artifacts at the inter-sections of adjacent image stacks.

4.4 Retrospectively ECG-Gated Multi-slice CT

Spiral CT acquisition with continuous table move-ment represented a very important step towards true volumetric imaging with mechanical CT. With spiral scanning, a true 3D data set can be acquired that consists of overlapping transaxial image slices (Kalender 1995). Additional advantages of spiral CT acquisition compared to sequential CT scan-ning are considerably improved spatial z-resolu-tion, faster scan speed, and larger volume coverage. ECG-triggered acquisition techniques are limited

to sequential scan modes and cannot be applied to spiral scanning.

Retrospectively ECG-gated spiral scanning is an attempt to synchronize the reconstruction of a continuous spiral scan to the movement of the heart by using an ECG trace that is recorded simultane-ously. The acquired scan data are selected for image reconstruction with respect to a pre-defined cardiac phase with a certain temporal relation to the onset of the R-waves that defines the start point of data used for image reconstruction.

Although retrospectively ECG-gated spiral scan-ning was introduced for sub-second single-slice spiral CT systems, its feasibility in clinical routine could not be demonstrated due to various limita-tions. To obtain the best temporal resolution, partial-scan-based reconstruction techniques are applied to the spiral scan data (Kachelriess 1998, Bahner 1999). However, no spiral interpolation algorithms could be used for retrospectively ECG-gated single-slice CT and data inconsistencies due to table move-ment and spiral artifacts were very common in transaxial slices. Slow table-feed is a precondition

Time

z - P

osit

ion

Scan

Scan

Scan

Scan

Scan

Scan

Scan

Scan

Fig. 4.5. Prospectively ECG-triggered scan coverage of the heart with a 16-slice CT scanner and 0.5-s rotation time. In the example, 16 slices with a total coverage of 16 mm (detector collimation 16 × 1 mm) are acquired with every individual scan. The entire scan range of 128 mm is scanned within the time of 15 heart beats, based on data acquired during eight heart beats. With the given heart rate of 60 bpm and the scan cycle time of about 1.5 s, scan data are acquired with every second heart beat

4.4 · Retrospectively ECG-Gated Multi-slice CT


Fig. 4.6a–d. Case examples acquired with prospectively ECG-triggered 4-slice CT with a 0.5-s rotation time (a, b) and 64-slice CT with a 0.33-s rotation time (c, d). With ECG-triggered 4-slice CT, 4 × 2.5-mm collimation, 120 kV, and 100 mA tube current, coronary calcifi cations in the circumfl ex and right coronary arteries can be detected with high sensitivity and high signal-to-noise ratio (arrows in a). Cardiac morphology and bypass patency can be evaluated with contrast-enhanced scanning using 4×2.5-mm collimation, 120 kV, and 200 mA tube current. In the example, a coronary bypass occlusion can be appreciated in the transaxial slices (arrow in b). With ECG-triggered 64-slice CT, 3- and 1.2-mm-thick slices can be generated from a 30 × 0.6-mm collimation setting. ECG-triggered 64-slice CT with 3-mm slices, 120 kV, and 126 mA tube current is used for coronary calcium quantifi cation (arrow in c). For contrast enhanced imaging of the cardiothoracic morphology, ECG-triggered 64-slice CT with 1.2-mm slices, 120 kV, and 250 mA can be used (d). Based on the fast rotation of 0.33 s and the underlying thin-slice scan ac-quisition with 0.6-mm collimation, 64-slice CT scanners have advantages in image resolution and image quality at higher heart rates. (Images courtesy of (a, d) University of Erlangen, Germany and (b,c) Klinikum Grosshadern, Munich, Germany)

a b

c d

for continuous and consistent volume coverage of the beating heart in all phases of the cardiac cycle, including diastole. Thus, single-slice systems hardly allow for continuous ECG-gated volume cover-age with overlapping slice acquisition of the entire heart and reasonable longitudinal resolution within reasonable scan times. Retrospective ECG gating can improve the robustness of ECG synchroniza-

tion against arrhythmia during the scan. However, single-slice acquisition produces a direct relation of the positions of the reconstructed slices and heart rate, including arrhythmia. For a given spiral table-feed, images with high overlap can be reconstructed for high heart rates; for low heart rates, gaps may be produced such that no fix reconstruction increment can be used.

81

Retrospectively ECG-gated multi-slice spiral scanning has the potential to provide an isotropic, 3D image data set of the complete cardiac volume without gaps and which can be acquired with a single breath-hold spiral acquisition. For ECG-gated reconstruction of scan data with continuous table-feed, dedicated multi-slice spiral reconstruc-tion algorithms are needed that are optimized with respect to temporal resolution, spiral artifact reduc-tion, and volume coverage. Such algorithms require certain pre-conditions of the scan data acquisition. Therefore, we will explain the technical principles of cardiac image reconstruction algorithms in the first section of this chapter before introducing the related acquisition techniques.

4.4.1 Multi-slice Cardiac Spiral Reconstruction

Multi-slice cardiac spiral reconstruction techniques allow for the reconstruction of overlapping images with fixed and heart-rate-independent image incre-ments at arbitrary z-positions and during any given phase of the cardiac cycle. These algorithms con-sist of two major steps and combine the previously described half-scan reconstruction for optimized temporal resolution with a multi-slice spiral weight-ing algorithm that compensates for table movement and provides well-defined slice sensitivity profiles. A representative example of these algorithms that also served as the basis for today’s algorithms is the so-called multi-slice cardiac volume reconstruction (MSCV) algorithm; (Ohnesorge 2000a), which is explained in detail below.

During the first step of the MSCV algorithm, so-called multi-slice spiral weighting, a “single-slice” partial-scan data segment is generated for each image using a partial rotation of the multi-slice spiral scan that covers the pertinent z-position. For each projection angle α within the multi-slice data segment, a linear interpolation is performed between the data of those two detector slices that are in closest proximity to the desired image plane zima. In contrast to standard multi-slice interpola-tion techniques (Klingenbeck 1999, Taguchi 1998, Hu 1999, Schaller 2000), each projection is treated independently. As a representative example, the

spiral interpolation scheme for a 4-slice CT system, including the calculation of the spiral interpolation weights for some representative projection angles, is illustrated in Figure 4.7. The presented approach is also used for today’s CT scanners with more than four slices and can easily be extended to 16- and 64-slice scanner geometry. With increasing scan time t and increasing projection angle α, the detector slices travel along the z-axis relative to the patient table. The z-position is normalized to the collimated slice width of one detector slice (SWcoll). Each multi-slice fan-beam projection Pf,M(αq,βm,n) consists of N sub-projections corresponding to the N detector slices that are measured at the same focus (source) position. (αq: projection angle of fan-beam projec-tion q, βm: the angle of a ray m within the fan relative to the central ray, n: detector slice n = 0,1,2,3 for the example of N = 4 detector slices). For reconstructing an image at a given z-position zima, the single-slice projections Pf(αq,βm) are calculated by linear inter-polation of those sub-projections within the multi-slice projection Pf,M(αq,βm,n) that are closest to the image z-position zima for a given projection angle αq (Eq. 4.3a). The interpolation weights are deter-mined according to the distances d(αq,n) of the sub-projections that are considered for reconstruction with z-positions zf(αq,n) to the image position zima (Eqs. 4.3b,c).

P w n P nf q m q f M q mn

n

( , ) ( , ) ( , , ),α β α α β= ⋅=

=

∑0

3

(4.3a)

w nd n d n SW

d n SWq

q q coll

q coll

( , )( , ) , ( , )

, ( , )α

α α

α=

− ≤

>

⎧⎨⎪

⎩⎪

1

0 (4.3b)

with

d n z z nq ima f q( , ) ( , )α α= − (4.3c)

A sole partial rotation of the multi-slice spiral scan can provide single-slice partial-scan data segments with linearly interpolated projections to reconstruct images that continuously cover a lim-ited z-range. That z-range can be extended if pro-jections of single-slice partial-scan data segments are generated with nearest-neighbor interpolation if two interpolation partners for linear interpolation are not available (Eq. 4.4).



PP z z

P z zf q mf M q m ima f q

f M q m ima f

( , )( , , ) , ( , )

( , , ) ,,

,

α βα β αα β

=≤≥

3 3

0 (( , )αq 0

⎧⎨⎩ (4.4)

Of course, nearest-neighbor interpolation may not be performed if the distances of the image z-position and the nearest measured sub-projection are too large. A reasonable limitation is to allow that a maximum of 50% of the projections within a single-slice partial-scan data segment may be gener-ated by using the projection of the nearest detector slice only. This situation is shown for image position zima,2 in Figure 4.7.

Special attention has to be paid to the slice sen-sitivity profiles (SSPs) as an important parameter of image quality in spiral CT. SSPs produced by the spiral weighting scheme of the MSCV algorithm have been evaluated according to measurements with a gold-plate phantom and based on theoretical calcu-lations in Flohr (1999). In this study, the SSPs and

resulting slice widths were assessed for different col-limated slice widths and different image positions. In a first-order approximation, MSCV provides a constant, pitch-independent relation (Eq. 4.5) of the collimated slice width (SWcoll) of one detector slice to the FWHM of the SSP representing the slice width (SW) of the reconstructed image.

SW 1.3 SWcoll (4.5)

Dependent on the weighting scheme that is present for a particular slice, the FWHM may vary between 1.1 SWcoll and 1.5 SWcoll (Flohr 2001). Slices that are generated in the center of the multi-slice partial scan (Fig. 4.7, zima,1) show symmetric profiles. How-ever, at boundary positions (Fig. 4.7, zima,2), where projections may be generated with nearest-neigh-bor interpolation, slices show moderate shifts of the center position on the order of 10%.

Partial Scan ≈ 260°

Partial Scan Single Slice Data

4-Slice Partial Scan Spiral Data

Interpolated Projection

zima,2

zima,1 d(α,n=1)

d(α,n=2)

Slice n=3

Slice n=2

Slice n=1

Slice n=0

Time t, Projection Angle α0

1

2

3

4

5

6z-

Posi

tion

Fig. 4.7. Generation of a single-slice partial-scan fan-beam data set from multi-slice fan-beam pro-jections acquired in a spiral scan with the MSCV algorithm. • Projections used for interpolation, ◊ interpolated projections. Continuous table movement is compensated for by linear interpolation within multi-slice fan beam projections in order to generate single-slice partial-scan data for image z-positions

83

MSCV also allows for retrospective generation of thicker slices for a certain collimated slice width SWcoll than given by Eq. 4.5, and it can be per-formed in a separate reconstruction using the same scan data. During the multi-slice spiral weighting step, up to three single-slice partial-scan data seg-ments are generated for each image z-position zima at closely adjacent z-positions zima–δz, zima and zima + δz (δz is a small distance in the z-direction) (Fig. 4.8). The weighted sum of these data segments before performing the partial-scan reconstruction step results in the reconstruction of a thicker slice at the desired position zima. Table 4.1 contains the appropriate parameters for reconstruction of the slice widths SW = 1.5SWcoll, SW = 2SWcoll, and SW = 3SWcoll. This technique is suited for reconstruction of images with reduced image noise for improved low-contrast resolution at the expense of reduced z-resolution.

Further advanced weighting schemes for multi-slice cardiac spiral reconstruction using 16- and 64-slice CT scanners have recently been devel-oped (Flohr 2003). In those generalized weight-ing schemes, spiral interpolation functions are used that involve more than only the two detector slices in closest proximity to the desired z-position zima. The slice width can be controlled by the shape of the interpolation function (Fig. 4.9). For recon-struction of wide slices relative to the collimation SWcoll, trapezoidal weighting functions are used that provide well-defined SSPs almost rectangular in shape, which help to reduce partial-volume arti-facts. The parameters a= 0.4, 0.5, and 0.6 determine the width of the weighting function (Flohr 2003). With edge-enhancing weighting functions, the usual slice-broadening by linear trapezoidal weight-ing functions can be avoided and slice widths can be generated that are equal to the collimated slice width SWcoll. Figure 4.10a, b shows the SSPs that

are generated for a 16-slice CT scanner for different heart rates and using different weighting functions. Different slice widths can be provided, and it can be demonstrated that the SSPs are largely indepen-dent from the heart rate. The smallest reconstructed slice width, measured by means of FWHM, using an edge-enhancing weighting function shows slice broadening of less than 10% compared to the colli-mated slice width. Based on the generalized weight-ing scheme, equivalent reconstructed slice widths can also be generated based on different collimation settings (Fig. 4.10c). However, SSPs that are based on thinner collimation are better defined and closer to the ideal rectangular shape, thus minimizing spiral artifacts.

Multi-slice cardiac spiral weighting generates partial-scan data segments for all required image z-positions zima in arbitrary image increments. Usu-ally, the selected increments are smaller that the reconstructed slice width in order to enhance z-axis resolution. The following, second step of the MSCV algorithm performs the previously described single-slice half-scan reconstruction (see Sect. 4.2) of the partial-scan fan-beam data at each image position zima. According to the sequential half-scan recon-struction, the temporal resolution in the center of the scan field of view equals half the rotation time. The time sensitivity profiles are spatially variable depending on the start and end positions of the tube for acquisition of the considered multi-slice partial-scan data set.

4.4.2 ECG-Gated Multi-slice Spiral Acquisition

For retrospectively ECG-gated reconstruction, each image is reconstructed using a multi-slice partial-scan data segment with an arbitrary temporal rela-

Table 4.1. Target positions for generating segments for the reconstruction of thicker slices

Slice width SW = 1.3SWcoll SW = 1.5SWcoll SW = 2SWcoll SW = 3SWcoll

z-positions for segments zima zima + 0.25 SWcoll zima + 0.5 SWcoll zima + SWcoll

zima – 0.25 SWcoll zima – 0.5 SWcoll zima

zima – SWcoll



Partial Scan ≈ 260°

Slice n=3

Slice n=2

Slice n=1

Slice n=0

Time t, Projection Angle α0

1

2

3

4

5

6z-

Posi

tion

zima

4-Slice Partial Scan Spiral Data

Slice zima-δz

Slice zima+δz

δz

δz

Fig. 4.8. Reconstruction of thicker slice-width with the MSCV algorithm by combining closely adja-cent single-slice partial-scan fan-beam data sets. The ratio between the collimation and slice thickness determines the distance and number of adjacent data sets that have be used

z-Distance in Factors of Collimated Slice-Width0

lin

trap 0.4trap 0.5

trap 0.6

edge enhance

1 2-0.5

0

0.5

1

Wei

gh

t Fa

cto

r

Fig. 4.9. Interpolation functions for a generalized cardiac spiral weighting approach. Trapezoidal weighting functions are used to generate wider slices with well-defi ned, rectan-gular-shaped slice sensitivity profi les (SSPs). Edge-enhancing weighting functions are used to generate thin slices with mini-mal slice-broadening compared to the collimated slice width

tion to the R-wave of the ECG trace. Image recon-struction during different heart phases is feasible by shifting the start point of image reconstruction rela-tive to the R-wave. For a given start position, a stack of images at different z-positions covering a small sub-volume of the heart can be reconstructed owing to multi-slice data acquisition. Figure 4.11 shows how the cardiac volume is successively covered with stacks of axial images (shaded stacks) reconstructed in consecutive heart cycles. Each stack that is recon-structed in one heart cycle covers a certain portion of cardiac anatomy. The maximum extension of these image stacks in the z-direction is determined by extension of the multi-slice detector, which can consist of from 4 to 64 individual detector rows. All image stacks are reconstructed at identical time points during the cardiac cycle. The detector rows travel along the z-axis relative to the patient table. The slope of the detector position lines represents

85

0

0.2

0.4

0.6

0.8

1

0-1-2-3 31 2z in mm

55 bpm FWHM = 0.831.051.522.013.04

0

0.2

0.4

0.6

0.8

1

0-1-2-4 41 3z in mm

FWHM = 3.043.00

Coll. 1.5 mm

Coll. 0.75 mm

-3 2

0

0.2

0.4

0.6

0.8

1

0-1-2-3 31 2

80 bpm

z in mm

Fig. 4.10a–c. Slice sensitivity profi les (SSPs) from a 16-slice CT scanner with 16 × 0.75-mm collimation using a generalized cardiac spiral weight-ing scheme. Different trapezoidal weighting functions allow slice widths between 1 and 3 mm to be generated. SSPs have been determined for a low heart rate of 55 bpm (a) and for a normal heart rate of 80 bpm (b). Both the calculated (lines) and measured (circles) results demonstrate the heart-rate independence of the SSPs. An edge-enhancing weighting function shows that the measured FWHM of the smallest reconstructed slice is 0.83 mm and that there is less than 10% slice-broading compared to the collimated slice width SWcoll = 0.75 mm. Based on the generalized weighting scheme, equivalent reconstructed slice widths can be gener-ated based on different collimation settings (c). However, SSPs based on thinner collimation are better defi ned and closer to the ideal rectangular shape, due to the use of trapezoidal weighting functions

a b

c

the pitch, which needs to be properly limited to allow for continuous volume coverage. In the shown example, an ECG gating approach was used with a delay that results in image reconstruction during the diastolic phase (shaded stacks). The hatched bars represent image stacks reconstructed from the same spiral data set, but in a different heart phase with a different delay parameter that, in turn, results in image reconstruction during the systolic phase. The entire heart volume can be reconstructed in both heart phases and in any other selected heart phase without gaps. Thus, a multi-phase reconstruction for true functional volume imaging of the moving heart can be generated from various 3D images with incrementally shifted delay parameters.

In each stack, single-slice partial-scan data seg-ments are generated equidistantly spaced in the z-direction depending on the selected image recon-struction increment. Continuous volume coverage

can only be achieved when the spiral pitch is appro-priately limited by the heart rate. In order to achieve full-volume coverage, the image stacks reconstructed in subsequent heart cycles must cover all z-posi-tions. If the pitch is too high, volume gaps between image stacks that are reconstructed using data from different heart cycles are present (Fig. 4.12). The pitch is limited by the patient’s heart cycle time (RR-interval time), as every z-position of the heart has to be covered by a detector slice at every time during one entire heart cycle. Consequently, the table may not move more than approximately the width of the multi-slice detector within one heartbeat. According to Eq. 4.6, the table feed is restricted to N-1 single slice widths of a detector with N detector slices within the time interval TRR + TQ, where TRR represents the RR-interval time for the present heart rate, Trot the full rotation time (between 330 ms and 500 ms), and TQ the partial scan time of a 240–260°



“3D” Image Data Trot / 2 Temp. Res.

Delay

R R R R

Reco

n ECG

Reco

n

Reco

n

Reco

n

Time

z - P

osit

ion Last Detector Row

First Detector Row

Detector Width

Continuous

Spiral Scan & Feed

Time

Reco

n

Reco

n

Reco

n

Reco

n

Reco

n

z - P

osit

ion

Spiral S

can

Reco

n

Fig. 4.11. a A scan with continuous table feed and continuous exposure is acquired for retrospec-tively ECG-gated multi-slice spiral scanning. b Stacks of overlapping images can be reconstructed with Trot/2 temporal resolution in every cardiac cycle with the MSCV algorithm. The individual stacks cover certain portions of the cardiac anatomy. Continuous 3D images can be reconstructed in different phases of the cardiac cycle by selection of the data ranges with certain phase relations to the R-waves

a

b

87

Reco

n

Reco

n

Reco

n

Reco

n

Reco

n

Time

z - P

osit

ion

Spiral S

can

Reco

n

Volume Gap

Volume Gap

Volume Gap

Volume Gap

Volume Gap

“3D” Volume Gaps

Delay

R R R RECG

Reco

n

Reco

n

Reco

n

Time

z - P

osit

ion

Last Detector Row

Detector Width

Continuous

Spiral Scan & Feed

First Detector Row

Fig. 4.12. a ECG-gated cardiac spiral examination with a multi-slice CT scanner at a pitch that is too high for the given heart rate. Continuous coverage of the heart volume is no longer possible. Volume gaps are present (hatched segments) between the image stacks that can be reconstructed in each cardiac cycle. b The gaps result in missing cardiac anatomy that could not be covered by the scan

b

a



partial rotation (e.g., TQ = 360 ms for Trot = 500 ms and TQ= 238 ms for Trot= 330 ms).

pitch [(N-1) Trot/(TRR+TQ)] / N (4.6)

N represents the number of illuminated detector slices, each with collimated slice width SWcoll. Based on Eq. 4.6, spiral weighting can be performed with two interpolation partners for all projections within the partial-scan data set if the interpolation function covers two detector widths (Fig. 4.9). If the interpola-tion function covers four detector widths (Fig. 4.9), at least three interpolation partners are available for all projections within the partial-scan data set. For example, with a heart rate of 60 bpm (TRR = 1000 ms), N = 16, Trot = 370 ms and TQ = 267 ms (e.g., SOMATOM Sensation 16, Siemens), Eq. 4.6 reveals a pitch limit of about 0.27. The maximum pitch depends on the number of illuminated detector slices and rotation time. With N = 32, Trot = 330 ms and TQ = 238 ms (e.g., SOMATOM Sensation 64, Siemens) the maxi-mum pitch for a heart rate of 60 bpm is reduced to 0.258. However, due to the larger width of the detector with N = 32 detector rows, a significantly increased volume coverage speed is still present compared to the scanner with N = 16 detector rows.

Scan speed can be increased for faster volume coverage by allowing that up to 50% of the interpo-lated projections are generated with nearest-neigh-bor interpolation at the edges of the image stacks. With this approach, the heart-rate-dependent spiral pitch is independent of the number of detector rows and is restricted according to Eq. 4.7. For example, with a heart rate of 60 bpm (TRR = 1000 ms), N = 16, Trot=370 ms, and TQ = 267 ms, the pitch can be increased from pitch = 0.27, according to Eq. 4.6, to pitch = 0.37, according to Eq. 4.7.

pitch Trot/TRR (4.7)

Scan protocols with pitch limitations according to Eq. 4.7 have been in use with 4-slice CT systems due to their limited volume coverage speed. With an increasing number of available detector rows (16 and even up to 64), most modern CT scanners use scan protocols with pitch limitations according to Eq. 4.6, due to significant image-quality advantages related to higher-quality SSPs.

The spiral pitch has to be selected prior to the scan and cannot be modified during the scan if the heart rate changes. Substantial drops in heart rate during the scan can produce local volume gaps. Therefore, the maximum heart cycle time (for the minimum heart rate) that is expected during the spiral scan should be used for calculation of the pitch for an individual patient. For modern 16- and 64-slice CT scanners, a spiral pitch of about 0.25 is feasible for most clinical applications, as the entire range of clinically relevant heart rates ≥45 bpm can be cov-ered and the heart volume can be scanned with the thinnest available slices within a single breath-hold (e.g., 12 cm in 15 s with N = 16, Trot = 370 ms, and SWcoll = 0.75 mm).

Retrospectively ECG-gated spiral scanning with 4-slice CT scanners (Ohnesorge 2000a) offered, for the first time, continuous volume imaging for quan-tification of coronary calcification (Fig. 4.13a) and high-resolution CT angiographic scanning of the coronary arteries (Fig. 4.13b). Moreover, assess-ment of cardiac function became possible by re-using the same scan data that had been acquired for evaluation of cardiac and coronary morphologies (Fig. 4.13c). The introduction of 16- and 64-slice CT technology and rotation times down to 0.33 s have significantly enhanced spatial and temporal resolu-tion as well as volume-coverage speed (Fig. 4.14), but the basic principles of cardiac scan acquisition and image reconstruction have remained the same since the appearance of the first 4-slice CT scanners. The significant improvement of spatial resolution from 4- to 16-slice and from 16- to 64-slice CT scanners is demonstrated in Figure 4.15. The application spectrum and clinical experience in cardiac and coronary imaging using the presented ECG-gated multi-slice spiral CT acquisition and reconstruction techniques will be presented separately, in detail, in the clinical application section.

4.4.3 Segmented Cardiac Reconstruction Algorithms

Conventional cardiac spiral reconstruction algo-rithms provide a continuous volume image of the heart with a temporal resolution equal to half the rotation time (Trot/2) of the individual slices. How-

89

LV

LALA

LV

Fig. 4.13a–c. Case examples obtained with retrospectively ECG-gated 4-slice spiral CT scans with 0.5-s rotation time. a Coro-nary calcifi cation of the left main and left descending coronary arteries can be identifi ed with high longitudinal resolution. Image reconstruction with overlapping increments enables more accurate evaluation of the calcifi ed plaque morphology in the z-direction. b The main coronary segments can be evaluated in a 3D reconstruction of a contrast-enhanced coronary CT angiography examination using volume-rendering technique. c Reconstruction of the same data set in end-diastole and end-systole allows for functional evaluation, e.g., of the left atrium (LA) and left ventricle (LV). Volume changes in the LA and LV and in mitral valve movement can be assessed. (Images courtesy of (a) University of Tübingen, Germany, and (b, c) Klinikum Grosshadern, Munich, Germany)

a b

c

ever, the resulting temporal resolution with today’s available minimum rotation times down to 330 ms may not be sufficient for motion-free imaging of the heart at high heart rates (Fig. 4.1) or for imaging the heart in phases with rapid cardiac motion. With unchanged rotation time, the temporal resolution can be improved by using scan data from more than one heart cycle for reconstruction of an image (“seg-

mented reconstruction”) (Lackner 1981, Kachel-riess 1998, Bruder 1999, Hu 2000, Kachelrieß 2000, Flohr 2001). The temporal resolution can be improved up to Trot/(2S) by using scan data of S sub-sequent heart cycles for image reconstruction, but at the expense of reduced volume-coverage speed or loss of resolution. To maintain good longitudinal resolution and thin-slice images, every z-position of



the heart has to be seen by a detector slice at every time during the S heart cycles. As a consequence, the larger the value of S and the lower the patient’s heart rate, the more the spiral pitch has to be reduced. If the pitch is too high, there will be z-positions that are not covered by a detector slice in the desired phase of the cardiac cycle. To obtain images at these z-positions, far-reaching interpolations have to be performed, which may degrade the SSP and reduce z-resolution (Kachelriess 1998, Kachelrieß 2000). If the selected spiral pitch is sufficiently small for continuous and gap-less volume coverage, the resulting increase in scan time usually needs to be

compensated by thicker slice collimation and thus reduced spatial z-resolution. Moreover, a reduction of the spiral pitch also correlates with a significant increase of radiation exposure. Improved temporal resolution alone, at the expense of reduced longitu-dinal resolution and increased radiation exposure, may degrade the overall diagnostic quality of stud-ies at low heart rates due to blurring of small ana-tomical details.

Heart-rate-adaptive algorithms were developed that provide adequately improved temporal reso-lution by using segmented reconstruction tech-niques that maintain high z-resolution at the same

Fig. 4.14a–c. Case examples obtained with retrospectively ECG-gated 64-slice spiral examinations. a The coronary artery tree can be visualized with signifi cantly increased spatial resolution and examined with sig-nifi cantly reduced breath-hold time and increased volume coverage compared to 4-slice CT. Reconstruction of the same data set in end-diastole and end-systole allows for functional evaluation of the cardiac chambers and of the cardiac valves. In the example, aortic-valve replacement can be visualized in the open (b, systole) and closed (c, diastole) positions. The reduced rotation time of 0.33 s allows for improved image quality also in systolic phase, when there is rapid cardiac motion. (Images courtesy of Jankharia Heart Scan Center, Bombay, India)

a

b

c

91

Scan direction

4 x 1 mm

16 x 0.75 mm

64 x 0.6 mm z-FFS

1.0 0.9 0.8 0.7 0.6 0.5 0.4 mm

a

b c

Fig. 4.15. a z-resolution phantom measurements for 4-, 16- and 64-slice CT scanners with ECG-gated spiral acquisition and reconstruction. The resting resolution phantom includes air-fi lled spheres 0.4–3.0 mm in diameter that can be visualized with multi-planar (MPR) cuts along the scan direction. The thinner collimation of 16-slice CT than of 4-slice CT provides an increase in z-resolution from about 0.9 mm with 4-slice CT to about 0.6 mm with 16-slice CT. Further reduced slice col-limation combined with double z-sampling technique (using a z-fl ying focal spot, z-FFS) enables a further increase of z-resolution for 64-slice CT to about 0.4 mm. Direct comparison of an examination of the same patient with 4-slice (b) and 64-slice (c) CT demonstrates the enhanced spatial resolution of 64-slice CT and its improved visualization of small cardiac structures, such as the fl aps of the aortic valve (arrowheads in b and c), and of complex coronary lesions (double arrows in b and c). Also, small-caliber distal coronary artery segments that could not be visualized with 4-slice CT (arrow in b) can be visualized with 64-slice CT (arrow in c)



time. A representative example of these algorithms is the adaptive cardiac volume (ACV) reconstruc-tion technique, which can use data from up to three consecutive heart cycles (Ohnesorge 2001, Flohr 2001). Depending on the patient’s heart rate during the scan, a variable number of heart cycles is used for image reconstruction, ranging from S = 1 heart cycle at low heart rates (temporal resolution Trot/2) up to S = 3 heart cycles at high pulse rates [tempo-ral resolution heart-rate-dependent up to Trot/(2S)]. With this adaptive approach, narrow SSPs may be obtained at adequate temporal resolution for a wide range of clinically relevant heart rates. A spiral pitch can be used that is sufficient to scan the complete heart with thin slices within a short single breath-hold and with only a moderate increase of radiation exposure.

If scan data are used for the reconstruction of an image that is acquired in S > 1 subsequent heart cycles, every z-position of the heart has to be cov-ered by a detector slice at every time during the S heart cycles. Thus, spiral pitch is not only limited by the heart rate but also by the number of heart cycles S used for reconstruction. If the table moves too quickly, interpolation between data acquired at a larger distance from the image plane can degrade the SSPs (Kachelriess 2000). As an extension of Eq. 4.7, the spiral pitch is limited by the minimum RR-inter-val time TRR during the scan, according to Eq. 4.8, in order to maintain a reasonable quality of the SSPs.

pitchN

NS

TT

rot

RR

≤ − +⎛⎝⎜

⎞⎠⎟

1 11 (4.8)

S is the number of subsequent heart cycles that are used for image reconstruction. The maximum pitch as a function of heart rate and number S of consecu-tive heart cycles used for reconstruction (S = 1, 2, 3) is shown in Fig. 4.16 for the example of a 16-slice CT scanner with 0.37-s rotation time. Reasonably fast volume coverage of the heart within one breath-hold and thin slices can be achieved with pitch = 0.2. According to Eq. 4.8, pitch = 0.2 implies that image reconstruction should be restricted to using data from S = 1 heart cycle only for moderate heart rates <63 bpm. For heart rates ≥3 bpm, S = 2 heart cycles may be used and S = 3 heart cycles for heart rates ≥95 bpm. For other pitch values, different heart-rate thresholds are implied.

Adaptive-segmented reconstruction algorithms automatically adapt the number of consecutive heart cycles used for image reconstruction accord-ing to the momentary heart rate of the patient during the scan (Fig. 4.17). These algorithms are exten-sions to single-segment methods that are limited to the use of one heart cycle only. Both major pro-cessing steps of single-segment algorithms, multi-slice spiral weighting and half-scan reconstruction, are also used for adaptive-segmented algorithms. During multi-slice spiral weighting, single-slice partial-scan data segments are generated from par-tial-scan data segments that are divided into S = 1, 2 or 3 sub-segments (depending on the heart rate), resulting in a temporal resolution of up to Trot/(2S). S = 1 sub-segment is used for low heart rates and reconstruction is performed with the conventional single-segment algorithm. At higher heart rates, the partial-scan data segment is divided into S = 2 or 3 sub-segments to improve temporal resolu-tion. Thus, multi-slice spiral data from two or three consecutive heart cycles, respectively, contributes to the single-slice partial-scan data segment. Each sub-segment is generated using data from one heart cycle only. Equivalent to single-segment algorithms, a linear interpolation is performed for each projec-tion angle αn within each sub-segment j (j = 0...S-1) between the data of those two detector slices that are in closest proximity to the desired image plane. The interpolation produces S single-slice sub-seg-ments located at the same z-position zima (Fig. 4.18) that can be assembled to a complete single-slice par-tial-scan data segment for image reconstruction. A certain overlap between the sub-segments can be taken into account to allow for smooth transition weighting, which reduces image artifacts due to data inconsistencies. For data consistency during image reconstruction, the sub-segments have to be picked at identical time points within the heart cycle, when the beating heart is exactly in the same relative phase position. The time points are determined by a certain distance relation to the R-waves. Data incon-sistency due to arrhythmic heart rates or irregular heart movement in consecutive heart cycles will result in degraded temporal resolution and substan-tial spatial blurring of moving structures.

Similar to the single-segment algorithms, the second step of adaptive-segmented algorithms per-

93

S=1

S=2

S=3

0.50

0.45

0.40

0.35

0.30

0.25

0.20

0.15

0.1040 50 60 70 80 90 100 120110

Heart Rate in BPM

Max

ium

um

Pit

ch

Fig. 4.16. Maximum spiral pitch as a function of heart rate and the number S (S = 1, 2, 3) of used data segments from consecu-tive cardiac cycles for the example of a 16-slice CT scanner with 0.37-s rotation time. Exceeding the spiral pitch limita-tion for a certain heart rate results in volume gaps. Minimum heart rates for use of S = 2 and S = 3 segments are shown for a pitch of 0.2, which allows for reasonably fast volume coverage with thin-slice collimation. Other pitch values imply different heart-rate thresholds. A pitch of 0.2 enables gap-less volume reconstruction for 16- and 64-slice CT scanners for all heart rates > 40 bpm

Delay

R R R R

Reco

n

Reco

n

Reco

n

Time

z - P

osit

ion

Last Detector Row

Detector WidthContinuous

Spiral Scan & Feed

Delay Delay

Low bpm High bpm High bpm

First Detector Row

S=1 SegmentTrot / 2

S=2 Segmentsup to Trot / 4

Fig. 4.17. Retrospectively ECG-gated multi-slice spiral reconstruction using adaptive-segmented reconstruction with up to S = 2 segments. Reconstruction with S = 1 segment (Trot/2 temporal resolution) is performed for low heart rates (e.g., ≤ 63 bpm). S = 2 segments (up to Trot/4 temporal resolution) are used for higher heart rates (e.g., > 63 bpm)

forms the previously described single-slice half-scan reconstruction of the partial-scan fan-beam data that are generated for each image position zima. The different data sub-segments from S individual heart cycles are transformed independent of the parallel-beam geometry and are appended as parallel-beam projections.

Adaptive-segmented reconstruction with S > 1 sub-segments only allows for improved temporal resolution if the patient’s heart rate and the rotation time of the scanner are appropriately de-synchro-nized. In a situation of optimal de-synchronization, the projection angles of the start- and end projec-tions of the sub-segments fit together and form a complete partial-scan data segment that contains 180° parallel projections after rebinning to paral-lel geometry (Fig. 4.19a). The partial-scan interval may then be divided into S sub-segments of equal size, and each sub-segment covers a temporal data interval Trot/4 for S = 2 or Trot/6 for S = 3 within the same relative heart phase. However, if heart rate and scanner rotation are synchronous, the same heart phase always corresponds to the same projection-angle segment, and a partial-scan interval cannot be divided into smaller sub-segments. As a conse-



Delay

R R

Time

z - P

osit

ion

Delay

Slice n=3

Subsegment j=0

Slice n=2

Slice n=1

Slice n=0Subsegment j=1

zima

Fig. 4.18. Generation of a single-slice partial-scan fan-beam data set from multi-slice fan-beam projections acquired in a spiral scan with adaptive-segmented reconstruction. • Projections used for interpolation, ◊ interpolated projections. Single-slice data segments are generated in consecutive cardiac cycles by linear interpolation at specifi c image z-positions. The individual data segments are combined into a single-slice partial-scan data set for image reconstruction at the considered z-position

quence, no better temporal resolution than Trot/2 is possible. In a situation with partially de-synchro-nized heart rate and rotation, the partial-scan inter-val may be divided into S sub-segments of different size, each covering temporal data intervals between Trot/2 and Trot/2S. The sub-segment with the longest temporal data interval then determines the tempo-ral resolution of the image (Fig. 4.19b). Segmented reconstruction techniques assume that all anatomy is exactly located in the same position in equivalent phases during the cardiac cycle. Slight shifts of the cardiac anatomy from one heart cycle to the next produce inconsistent scan data and may result in image artifacts (Fig. 4.20). Generally, adaptive-seg-mented reconstruction algorithms generate images with temporal resolution in the interval [Trot/2,

Trot/2S], depending on the relation of rotation time and patient heart rate. During the same scan, tempo-ral resolution can vary with changing heart rate as the rotation time is fixed. The relationship between temporal resolution and rotation time and heart rate is demonstrated in Figure 4.21 for different rotation times between 0.5 s and 0.33 s and up to S = 2 seg-ments. Figure 4.22 shows the temporal resolution as a function of heart rate for a 64-slice CT scanner with 0.33-s rotation time and up to S = 4 segments. The temporal resolution is calculated as the FWHM of the time sensitivity profiles in the center of the scan field of view. Although excellent temporal resolution can be achieved for certain heart rates, for others the heart cycle and scanner rotation are perfectly synchronized and the temporal resolution

95

90°

90°

X°

90°

90°

Delay

R R R R

Delay Delay Delay

90° 90° 90° 90°+X°

Subsegment j=0 Subsegment j=0

Subsegment j=1 Subsegment j=1

Proection Angle αn

Proection Angle αn

Time t, Projection Angle αn Time t, Projection Angle αn

Fig. 4.19. a For segmented reconstruction with S = 2 segments and optimal de-synchronization of heart rate and system rotation, the sub-segments form a complete partial-scan data segment without overlap and with temporal resolution Trot/4. b For only partially de-synchronized heart rate and system rotation, the sub-segment with the longest temporal data interval determines the temporal resolution of the image

1st2nd3rd

Fig. 4.20a, b. Segmented reconstruction with S = 3 segments. The segments have different temporal widths for the given rotation time and heart rate. a Sub-segment 1, with the longest temporal data interval, determines the temporal resolution of the image. However, the coronary anatomy may be located at slightly shifted positions during consecutive heart cycles, which may result in data inconsistencies and blurring artifacts. b In the example, coronary angiography shows the slightly shifted location of the center line of the right coronary artery in three consecutive heart cycles

a b


a b


0.5s

0.42s

0.37s

0.33s

125–250ms

105–210ms

83–165ms

92–185ms

1 Segment 2 Segments

Heart Rate

Tem

po

ral R

eso

luti

on

[ms]

250

200

150

100

50

50 60 70 80 90 100

Fig. 4.21. Demonstration of temporal resolution as a function of heart rate and rotation time for adaptive-segmented reconstruction using up to S = 2 segments. The heart rate threshold for switching from S = 1 to S = 2 segments is set to 63 bpm. Tem-poral resolution is strongly heart rate dependent within the interval [Trot/2, Trot/2S]. The very differ-ent locations of maxima and minima for different rotation speeds suggest that heart-rate-dependent selection of rotation speed might be feasible only in patients with stable and predictable heart rate. Reconstruction with S = 1 segment can also be used for higher heart rates. The temporal resolu-tion then remains constant at Trot/2, independent of heart rate

40

60

80

100

120

140

160

40 60 80 100 125 140

65 b

pm

82 b

pm

105

bp

m

Tem

po

ral R

eso

luti

on

[ms]

Heart Rate / bpm

40

60

80

100

120

140

160

50 65 75 100 125 150

Tem

po

ral R

eso

luti

on

[ms]

Heart Rate / bpm

1,2-segment3-segment4-segment

Fig. 4.22. Temporal resolution as a function of heart rate for adaptive-segmented reconstruction with up to S = 4 segments using a 0.33-s rotation time. With S = 4, temporal resolution can be increased to Trot/8 only for certain heart rates. With S = 4 segments, temporal resolution changes signifi cantly with only small changes in heart rate. The locations of maxima and minima are strongly dependent on rotation time. Temporal resolution may be diffi cult to predict in clinical conditions – given the very strong variation of temporal resolution with heart rate and rotation speed – when using S = 3, or S = 4 segments for reconstruction

of Trot/2 may not be improved by using data from multiple heart cycles. Image quality improvements with adaptive-segmented reconstruction techniques can be demonstrated in patients with higher heart rates. A direct comparison of single-segment recon-struction and reconstruction with S = 2 segments can reveal significantly improved delineation of coronary calcifications in a patient with a heart rate of 76 bpm (Fig. 4.23).

Image quality and resolution along the scan direction provided by segmented reconstruction

algorithms have been assessed with a longitudinal resolution phantom that included air-filled spheres with diameters ranging from 0.4 to 3 mm (Flohr 2001, Flohr 2003). It could be demonstrated that adaptive-segmented reconstruction can provide stable longitudinal resolution without a consider-able influence of heart rate (Fig. 4.24).

The possible image quality improvements of segmented reconstruction algorithms compared to single-segment reconstruction can be demonstrated in examinations of patients with high heart rates

97

0.37s, 76bpm, 1-Segment 0.37s, 76bpm, 2-Segments

1009080706050Heart Rate

1009080706050Heart Rate

50

150

200

250

100

50

150

200

250

100

[ms]

[ms]

0.37s 0.37s185ms

130ms

1 Segment 1 Segment 2 Segments

Fig. 4.23. Case example of a 16-slice CT cardiac examination that demonstrates the improvement of image quality with adaptive segmented vs. single-segment reconstruction at higher heart rate. A coronary calcium scan was performed in a patient with 76 bpm during the scan with: 16 × 1.5-mm collimation, 0.37-s rotation, reconstruction in diastole at 55% of the RR-interval. Coronary calcifi cations in the left anterior descending and right coronary arteries (arrows) are depicted with considerably less motion using adaptive-segmented reconstruction (right, S = 2 segments, 130-ms temporal resolution) compared to single-segment reconstruction (left, S = 1 segment, 185-ms temporal resolution). (Case courtesy of Tübingen University, Germany)

(Fig. 4.25) or for reconstruction of images in phases of the cardiac cycle with rapid cardiac motion (Fig. 4.26). However, cardiac anatomy can be located at slightly shifted positions even in equivalent phases during the cardiac cycle. Due to these incon-sistencies, segmented reconstruction techniques may produce degraded image quality compared to single-segment reconstruction, despite higher tem-poral resolution (Fig. 4.27). To achieve most robust cardiac image quality with the lowest possible radiation exposure, single-segment reconstruction techniques combined with fast rotation speeds are usually preferred in clinical practice. The use of seg-mented reconstruction techniques is limited to spe-cial clinical situations, such as patients with either very high heart rates or stable and predictable heart rates, and for reconstruction of the larger cardiac anatomy during phases with rapid cardiac motion in order to analyze cardiac function.

4.4.4 Cardiac Cone-Beam Reconstruction Algorithms

The cardiac image reconstruction algorithms intro-duced thus far were designed for multi-slice CT sys-tems with only a limited number of detector rows, and did not take into account the conical shape of the X-ray beam in the axial direction. As demonstrated in Chapter 3, cone-beam reconstruction algorithms are mandatory for general-purpose CT scanning with eight and more detector slices to avoid severe image artifacts. The severity of cone-beam-induced artifacts depends on the number of simultaneously acquired slices, on the width of each slice, and on the distance of a considered object from the center of the scan field of view. Cone-beam artifacts are most pronounced at high contrast boundaries; thus, typical sources of cone-beam artifacts are the ribs, pelvic bones, and bronchi. Since the heart is



Fig. 4.25a–c. Case examples of a 64-slice CT scanner with 0.6-mm collimation and 0.33-s rotation time. The fi gure shows the improved image quality of cardiac and coronary CT an-giography examinations with adaptive-segmented vs. single-segment reconstruction at higher heart rate. The patient’s heart rate was between 106 and 110 bpm during the scan. Images were reconstructed at 35% of the RR-interval. Motion artifacts are present in the images based on single-segment reconstruction with 165-ms temporal resolution (a, arrows). Two segments can increase temporal resolution from 165 ms to about 90 ms using adaptive segmented reconstruction, and motion artifacts can be largely eliminated (b). Three segments allow a further improvement in temporal resolution to 70 ms, but image quality is compromised due to the combination of data from three consecutive heart beats (c). (Cases courtesy of Klinikum Grosshadern, University of Munich, Germany)

Phantom

55 bpm 1 Segment

70 bpm 2 Segments

100 bpm 3 Segments

0.4 0.90.80.6 1.0 3.0 mm1.2 1.5 2.0

0.4 0.90.80.6 1.0mm0.70.5

55 bpm 1 Segment

75 bpm 2 Segments

90 bpm 2 Segments

Fig. 4.24a, b. Investigation of the infl uence of heart rate on im-age quality and resolution in the scan direction with 16-slice (a) and 64-slice (b) CT using a resting longitudinal resolution phantom (see Figs. 4.3 and 4.15) and simulated ECG signals with different heart rates that control the adaptive-segment-ed reconstruction. The 16-slice CT scans were acquired with 0.75-mm collimation, 0.37-s rotation, pitch 0.2, 0.75-mm slice width, and 0.5-mm increments. A longitudinal resolution of 0.6 mm can be achieved for all anticipated heart rates. The 64-slice CT scans were acquired with 0.6-mm collimation (us-ing double z-sampling technique), 0.33-s rotation, pitch 0.2, 0.6-mm slice width, and 0.3-mm increments. A longitudinal resolution of 0.4-mm can be achieved for all anticipated heart rates. In 16-slice and 64-slice CT phantom tests, image quality is largely independent of both heart rate and the number of segments used for reconstruction

a

b

usually centered and does not contain large high-contrast structures, cone-beam artifacts are largely negligible for CT scanners up to 16 slices (Flohr 2003). However, cardiac cone-beam reconstruction algorithms gain importance for CT scanners with a higher number (32 or 64) of detector slices as well as for ECG-gated examinations of the chest and diag-nosis of anatomy that is located at a larger distance from the center of the scan field of view. The cardiac cone-beam reconstruction algorithms in use today (Bruder 2001, Bruder 2002) represent extensions of those in use for general-purpose multi-slice spiral reconstruction and have been successfully tested for the latest multi-slice CT scanners with 32 and up to 64 detector slices.

Similar to cone-beam reconstruction techniques in multi-slice body CT imaging, cardiac cone-beam reconstruction techniques generate double-oblique image stacks (so-called booklets) that are individu-ally adapted and optimally fitted to the spiral path. In a second step, these booklets are reformatted to a set of overlapping trans-axial images (Fig. 4.28). For single-segment reconstruction, one booklet is used for reconstruction and reformation in each heart cycle. Data segments that cover 180° of parallel-beam projections are employed for the generation of these booklets, leading to a temporal resolution of the images within the booklet of half of the rota-tion time of the scanner. In the case of segmented

99

ba c

Fig. 4.26a–d. Case example of a cardiac CT angiography examination for a pa-tient with a heart rate of 100–107 bpm using a 64-slice CT scanner with 64 × 0.6-mm slices and 0.33-s rotation time. Images are reconstructed during the systolic phase with a relative delay of 25% of the RR-interval, and MPRs are generated in the short heart axis. With single-segment reconstruction and a temporal resolution of 165-ms, motion artifacts are present (a, arrows). Adap-tive-segmented reconstruction with 2 (b), 3 (c), and 4 (d) segments improves temporal resolution up to 60 ms, and cardiac anatomy can be visualized dur-ing phases of rapid cardiac motion. A substantial image quality improvement takes place when using two segments instead of one. Further image qual-ity improvements are only minor when using more than two segments. (Cases courtesy of Klinikum Grosshadern, Uni-versity of Munich, Germany)

b

d

a

c



Scan Time [s]

Heart Rate [bpm]

Scan Start Scan End

0 2 4 6 8 10 12

69 95 71 90 77 84 86 87 89 87 86 85 84 83 83 84

Fig. 4.27a–e. Case example of a cardiac CT angiography using a 64-slice CT scanner with 64 × 0.6-mm slices and 0.33-s rotation time. a The patient had a varying heart rate between 69 and 95 bpm during the scan. The arrows in a point to the scan start and to the scan end, and illustrate the scan time interval of about 11 s. b, d Images were reconstructed with a relative delay of 45% of the RR-interval using single-segment reconstruction with 165-ms temporal resolution. c, e Three-segment reconstruction with increased temporal resolution between 55 and 165 ms depending on heart rate. Despite the higher effective temporal resolution of 3-segment reconstruction, motion artifacts are still present (arrows). In some segments, even better image quality is achieved with single-segment reconstruction (double arrows, triple arrows) due to beat-to-beat data inconsistencies with 3-seg-ment reconstruction in the presence of heart-rate changes. (Cases courtesy of Klinikum Grosshad-ern, University of Munich, Germany)

a

b c

d e

101

Fig. 4.28. The principle of cardiac cone-beam reconstruction with generation of individual double-oblique image stacks (so-called booklets) in each heart cycle that are subsequently reformatted to overlapping transaxial image slices

cardiac reconstruction data, segments from con-secutive heart cycles that cover less than 180° of parallel-beam projections are used for image forma-tion. Segmented cardiac cone-beam reconstruction algorithms permit reconstruction of image booklets from data segments covering less than 180°; how-ever, these do not represent complete CT images due to truncation of the projection segments. The incomplete CT images from individual heart cycles are separately reformatted and subsequently com-bined to complete axial CT images containing con-tributions from multiple consecutive heart cycles that add up to a total parallel projection interval of 180°.

The elimination of cone-beam artifacts in car-diac CT images has been demonstrated in phantom studies using data from multi-slice CT scanners with up to 64 individual detector rows (Bruder 2002) (Fig. 4.29). Cardiac cone-beam reconstruction algorithms have also demonstrated considerable image quality improvements in ECG-gated cardiac studies with latest 64-slice CT scanners, in particu-lar for the diagnosis of peripheral thoracic anatomy located further from the center of the scan field of view (Fig. 4.30).

4.4.5 ECG-Gated Spiral Scanning with Increased Volume Speed

Both ECG-triggered sequential scanning and ECG-gated spiral scanning require substantially slower scan speeds than conventional multi-slice spiral scanning without ECG-gating. Normal ECG-trig-gered or ECG-gated scan techniques using the pres-ent 4- and 16-slice CT scanners cannot provide thin-slice coverage of the entire chest anatomy with a scan range of about 350 mm within a short single breath-hold. However, thoracic CT studies are frequently degraded by motion artifacts caused by transmitted cardiac pulsation (Loubeyre 1997, Schöpf 2000) and elimination of motion artifacts via ECG syn-chronization may substantially improve diagnostic quality. Although the latest 32- and 64-slice CT scan-ners are capable of covering the entire chest with ECG-gated scan protocols in about 15–20 s, a further reduction of breath-hold time remains feasible.

Scan speed can be increased with a modified ECG-gated spiral acquisition technique that enables image reconstruction with high temporal resolution by the use of higher pitch values (Flohr 2002). The



Fig. 4.29a–d. Evaluation of cardiac cone-beam reconstruction algorithms using an anthropomor-phic heart phantom. CT data were generated for a multi-slice CT scanner with 64 × 0.625-mm col-limation (40-mm coverage per rotation, no z-fl ying focal spot) and reconstructed using cardiac reconstruction algorithms without (a, c) and with (b, d) cone-beam correction. Due to the larger volume coverage per rotation, images without cardiac cone-beam reconstruction demonstrate sig-nifi cant cone-beam artifacts in the heart and in the periphery. These can be eliminated using cardiac cone-beam reconstruction

a b

c d

limitation of the spiral pitch for ECG-gated multi-slice spiral scanning results is a consequence of the phase-consistent coverage of the heart volume, i.e., data may only be used during identical relative time points within the cardiac cycle. The spiral pitch can be increased by allowing images that are part of the same continuous volume to be reconstructed with half-scan reconstruction algorithms but in differ-ent phases of the cardiac cycle. In this approach, data have to be excluded from image reconstruction acquired during phases of high cardiac motion (i.e., systole). The data windows that are excluded from reconstruction during the different cardiac cycles may be positioned relative to the onset of the R-waves

with a certain temporal relation (Fig. 4.31). A stack of images with a given increment is reconstructed during each cardiac cycle and combined to form a volume image data set. For each image, an interpo-lated partial-scan data segment is generated with the previously described single-segment cardiac spiral reconstruction approach. Thus, the temporal resolution of each individual image is equal to half of the rotation time. The data segments selected for reconstruction of individual slices are shifted equi-distantly in time within the heart cycle. Pulsation artifacts are usually most severe during the systolic phase of cardiac contraction. Therefore, to eliminate image degradation due to systolic pulsation, scan


Fig. 4.30a–f. Demonstration of image-quality improvements using cardiac cone-beam reconstruction techniques with a 64-slice CT scanner. CT data were acquired in a chest phantom for a 64-slice CT scanner with 32 × 0.6-mm collimation (19.2-mm coverage per rotation, with z-fl ying focal spot) and reconstructed using cardiac reconstruction algorithms without (a, c, e) and with (b, d, f) cone-beam correction. Due to the limited volume coverage per rotation, images can be reconstructed without cone-beam reconstruction to display the cardiac anatomy, and cone-beam artifacts are negligible (arrows in a and b). However, such artifacts are present in the periphery of the scan fi eld of view (arrows in c and e) and cardiac cone-beam reconstruction is required to reduce artifacts, in particular for the peripheral thoracic anatomy (arrows in d and f)

a b

c d

e f


Time

Delay

Skip

Detector Width

z - P

osit

ion

Skip

Skip

“3D” Image Data Trot / 2 Temp. Res.

∆TS

Reco

nContinuous

Spiral Scan & Feed

Reco

n

Reco

n

Last Detector Row

First Detector Row

Fig. 4.31. Volume coverage with fast ECG-gated spiral acquisition using increased pitch. Stacks of overlapping images can be reconstructed with Trot/2 temporal resolution in every cardiac cycle. Data ranges acquired during phases with fastest cardiac motion are omitted for reconstruction of 3D volume image data. The omitted temporal range is defi ned with an adjustable width ∆TS and an adjustable phase related to the onset of the R-wave

data acquired during this phase of cardiac contrac-tion is omitted within a temporal window of fixed width ∆TS. This window is defined within the car-diac cycle with a fixed temporal relation to the onset of the R-wave. Continuous volume coverage requires that the image stacks that are reconstructed in con-secutive heart cycles overlap in the z-direction. Thus, the spiral pitch is restricted to a value given by the fixed temporal width ∆TS of the data window that is omitted during image reconstruction (Eq. 4.9). Here, the pitch limitation is independent of heart rate.

pitchN

NT

T Trot

S rot

≤ − ⋅+

1

2∆ / (4.9)

N represents the number of slices and Trot is the scanner rotation time. It has been shown that values between ∆TS 400 ms and ∆TS 500 ms are a reason-able approximation for the duration of the phase of strongest cardiac pulsation. The resulting pitch values between 0.5 and 0.6 allow the entire thorax to be covered with thin slices within a single breath-hold. As an example, a modern 16-slice CT scanner can cover a 350-mm scan range with 16 × 0.75-mm

collimation, 0.37-s rotation, and a pitch of 0.5 in a breath-hold time of 22 s. Examples of examinations of the thoracic aorta are shown in Figs. 4.32 and 4.33. Direct comparison of conventional non-gated spiral reconstruction and ECG-gated reconstruction of the same scan data demonstrates a substantial improve-ment with the latter technique, since motion arti-facts can be largely eliminated.

With the advent of 32- and 64-slice CT scanners, thin-slice ECG-gated coverage of the entire chest has also become feasible with regular ECG-gated car-diac scan protocols. Based on a usual pitch value of 0.2, which can provide gap-less volume coverage for a heart rate down to about 40 bpm, a 350-mm scan range can be covered within a breath-hold time of 10–20 s. For patients with heart rate of 60 bpm and above during the scan, the pitch can be increased to values of about 0.3, thus providing a 50% increase in volume coverage speed. Cardiac cone-beam recon-struction algorithms become increasingly impor-tant for ECG-gated scanning of the cardiothoracic anatomy with 32- and 64-slice CT scanners, since large portions of the thoracic anatomy are located

105

Fig. 4.32a, b. Fast ECG-gated spiral acquisition with a 4-slice CT scanner for evaluation of the thoracic aorta: comparison with multi-slice spiral reconstruction using a standard algorithm. a With standard reconstruc-tion, considerable pulsation artifacts that produce a double contour are visible. b The intimal fl ap of a type-A dissection can be evaluated free of pulsation artifacts on a corresponding transverse section using fast ECG gating. Scan and reconstruction parameters: 4 × 1-mm collimation, 0.5-s rotation, 120 kV, 300 mA/200 mAs, pitch 0.6, 1.25-mm slice-width, 0.8-mm image increments, window of omitted data has width ∆TS = 400 ms and starts 0 ms after onset of the R-wave. (Images courtesy of the University of Tübingen, Germany)

a b

Fig. 4.33a, b. Case study of a CT angi-ography examination of the thoracic and abdominal aorta using a 16-slice CT scanner with 16 × 0.75-mm collimation and 0.37-s rotation time and fast ECG gating. In the conventional non-gated technique, the displayed scan range of 500 mm can be scanned in 10 s with a pitch of 1.5 but pulsation artifacts may be present. With a fast ECG-gated spiral protocol and pitch 0.5, the scan time is increased to 30 s but pulsation artifacts are largely eliminated. In the presented case, fast ECG-gated acquisition demon-strates an aortic dissection (type B) free of motion over its entire course in sagit-tal MPR (a) and volume-rendering tech-nique (b) (Case by courtesy of University Hospital Graz, Austria) ba



at greater distance to the center of the scan field of view and thus are subject to cone-beam artifacts.

4.5 Synchronization with the ECG and Cardiac Motion

4.5.1 ECG-Based Phase Selection

With both prospective ECG triggering and retro-spective ECG gating, the starting points of data acquisition or the start points of data selection for reconstruction have to be defined within each car-diac cycle during the acquisition. These start points are determined relative to the R-waves of the ECG-signal by a phase parameter. The following phase-selection strategies can be used (Fig. 4.34).

• Relative delay: A temporal delay Tdel relative to the onset of the previous R-wave is used for determining the start point of the ECG-triggered acquisition or the start point of the reconstruc-tion data interval. The delay time Tdel is deter-mined individually for each heart cycle as a given percentage δRR of the RR-interval time TRR. For ECG triggering, the RR-interval times have to be prospectively estimated based on the prior RR-interval times.

• Absolute delay: Fixed delay times Tdel after onset of the R-wave defi ne the start point of the ECG-triggered acquisition or the start point of the reconstruction data interval.

• Absolute reverse: Fixed times Trev prior to the onset of the next R-wave defi ne the start point of the ECG-triggered acquisition or the start point of the reconstruction data interval. For ECG trig-gering, the position of the next R-wave has to be prospectively estimated based on the prior RR-interval times.

Fig. 4.34a–c. Phase defi nition for ECG triggering and ECG gating by selection of the start point of the temporal data interval within every heart cycle. The following different phase-defi nition strategies are used. a Relative delay: delay time after the previous R-wave, determined as a fraction δRR of the RR-interval. b Absolute reverse: constant interval Trev prior to the next R-wave. c Absolute delay: constant delay time Tdel after the previous R-waveTime

Time

Time

TRR,1 TRR,2

TRR,1 TRR,2

TRR,1 TRR,2

R

ECG

R R

R R R

R R R

ECG

ECG

Tdel TdelAbsolute Delay

Absolute Reverse

Trev Trev

RelativeDelay

Tdel,1=δRR TRR,1 Tdel,2=δRR TRR,2Image Time

Interval

a

c

b

107

Different approaches are in use in clinical prac-tice today depending on the clinical application. For motion-free imaging of small anatomical struc-tures (i.e., coronary arteries) in diastolic phase, with least cardiac motion, the relative delay and absolute reverse approaches are most frequently used. Car-diac image quality greatly depends on the cardiac phase selected for image reconstruction (Fig. 4.35), as cardiac motion varies widely during different phases within the cardiac cycle (Fig. 4.1). Cardiac image reconstruction is usually performed during phases of least cardiac motion – usually between mid-diastole and end-diastole of the cardiac cycle. Within the ECG trace, end-diastole is usually rep-resented by the onset of the P-wave. For the latest 64-slice CT scanners, which provide higher tem-poral resolution, cardiac image reconstruction can also be feasible during end-systole. The latter rep-resents a shorter phase of low cardiac motion and is marked within the ECG trace by the onset of the T-wave. Despite intensive research to standardize phase selection for cardiac image reconstruction, heart-rate-dependent and patient-individualized optimization usually remain necessary to obtain the best possible results (Hong 2001a, Kopp 2002).

For functional imaging with retrospective ECG gating, images need to be reconstructed in phases of maximum and minimum filling of the ventri-cles (end-diastole and end-systole). End-diastolic reconstruction is feasible with the absolute reverse approach, while the absolute delay approach allows for most consistent reconstruction in end-systolic phase.

A regular heart rate without significant and sudden heart rate changes during the scan is usu-ally a pre-requisite for diagnostic image quality for examinations of small cardiac and coronary anat-omy. A strongly irregular heart rate may result in substantial data mis-registration and image arti-facts, thus leading to results that cannot be used diagnostically (Fig. 4.36). However, retrospective ECG gating allows for viewing and analysis of the ECG signal after the end of the scan, and data are available during all phases of the cardiac cycle. This offers the possibility of retrospectively modifying synchronization of the ECG trace and data recon-struction. Interactive editing of R-peak positions that are detected inappropriately or represent irregular heart beats can have a positive impact on phase con-sistency and image quality in patients with irregular

0% 10% 20% 30% 40%

50% 60% 70% 80% 90%

Fig. 4.35. Cardiac image reconstruction of data acquired with a 64-slice CT scanner with 0.33-s rotation time in different phases of the cardiac cycle using the relative-delay approach. Images were reconstructed with single-segment reconstruction, thus providing a temporal resolution of 165 ms. Image quality largely depends on the selected phase of the cardiac cycle. The best image quality can be obtained in mid-diastolic phase (δRR = 50%) and end-systolic phase (δRR = 30%)

4.5 · Synchronization with the ECG and Cardiac Motion


heart beats (Fig. 4.37). Also, individual adjustment of the image-time interval positions – independent of the position that is determined by the ECG-gating parameter – may be useful in patients with substan-tial arrhythmias.

4.5.2 The Pros and Cons of ECG Gating and ECG Triggering

Retrospectively ECG-gated spiral scanning with single-slice CT systems featuring sub-second rotation has been tested for coronary artery and cardiac-func-tion imaging in clinical trials but serious limitations have been discovered (Mochizuki 2000). With the advent of multi-slice acquisition, ECG-gated spiral scanning has become feasible, and with significant advantages over prospective ECG triggering that are important for clinical applications.

• ECG-gated spiral scanning provides continuous volume coverage and better spatial resolution in the patients' longitudinal direction, as images can be reconstructed with arbitrary, overlapping slice increments. ECG-triggered sequential scan-ning is usually restricted to scanning with non-overlapping adjacent slices or slice increments with only small overlap. The scan time to cover the heart volume is thus directly proportional to the slice increment.

• Retrospective analysis of the ECG results in less sensitivity to heart-rate changes during the scan. The ECG trace can be retrospectively analyzed and extra-systolic beats can be eliminated for reconstruction. With prospective ECG trigger-ing, estimation of the next RR-interval may be incorrect when heart-rate changes are present (e.g., arrhythmia, Vasalva maneuver) and scans may be placed in inconsistent heart phases. In that case, even slices with high temporal resolu-

bpm 64 94 61 98 91 63 93 90 88 60

Fig. 4.36. Case example of a patient scanned with a 16-slice CT scanner with 16 × 0.75-mm collimation and 0.42-s rota-tion time. The patient had a strongly irregular heart rate that changed between 60 and 98 bpm during the scan. Although the quality of the individual image stacks reconstructed from individual heart beats is acceptable, the image quality of 3D reformats is heavily compromised due to misalignment of the image stacks for consecutive cardiac cycles. (Case courtesy of Münster University, Germany)

109

tion that are free of motion artifacts cannot be used for continuous 3D image data sets.

• ECG-gated spiral scanning provides faster volume coverage than ECG-triggered sequential scanning because spiral scan data can be acquired continu-ously and images can be reconstructed in every cardiac cycle. Relatively long travel distances and travel times of the table are present for multi-slice acquisition in between two consecutive scans. This limits the scan-cycle time (minimum time in between the start of two consecutive scans), and ECG-triggered scans can often be obtained only in every second heart beat for patients with higher heart rates.

• ECG-gated spiral acquisition allows for imag-ing in a complete cardiac cycle using the same scan data set, thus providing information on

cardiac function. ECG-triggered acquisition tar-gets only one specifi c phase of the cardiac cycle and requires additional examinations with new breath-hold levels and additional contrast agent to cover more phases of the cardiac cycle.

During ECG-gated spiral imaging of the heart, data are acquired with small spiral pitch (pitch << number of slices, i.e., overlapping acquisition) and continuous X-ray exposure. Thus, ECG-gated spiral acquisition requires a higher patient dose of radia-tion than ECG-triggered sequential acquisition for comparable signal-to-noise ratio. All spiral data can be used for image reconstruction in differ-ent cardiac phases and no data have to be omitted. However, if only one dedicated cardiac phase (i.e., diastolic phase) needs to be targeted by retrospec-

76 90 54 99 40

19 20 21 22 23

76 90 59 100 51

19 20 21 22 23t

Fig. 4.37a, b. Case example of a patient scanned with a 64-slice CT scanner. The patient had extra-systolic beats during the scan. The scan was acquired using a slice acquisition of 64 × 0.6 mm and a rotation time of 0.33 s. Misregistration artifacts can be strongly reduced and the diagnostic quality of the study restored via interactive elimination of the extra heart beats before image reconstruction. In this case, an extra-systolic beat strongly compromises visualization of the distal right coronary artery (a). After interactive elimination of those reconstruction intervals affected by the extra-systolic beat (blue bars) and manual introduction of alternative image reconstruction intervals (gray bars, green arrow), image quality could be fully restored (b). (Case courtesy of New York University, USA)

a b



tive data selection, the specific requirements of the clinical application should indicate whether ECG-triggered sequential scanning with less radiation exposure could provide sufficient performance and image quality. Newer developments enable a reduc-tion of radiation exposure during retrospectively ECG-gated scanning. The goal of these techniques is to maintain the important benefits of ECG-gated spiral scanning but to reduce X-ray radiation expo-sure to levels comparable to those of ECG-triggered sequential acquisition, as will be explained in a later chapter.

4.5.3 Alternative Cardiac-Motion Gating Approaches

Phase selection of ECG-synchronized CT scans is usually performed with respect to the temporal position of the R-waves. The R-wave can be easily detected from the patient's ECG due to its high signal amplitudes. Cardiac image reconstruction is frequently carried out during end-systole and end-diastole, which are the phases of the cardiac cycle with the least cardiac motion and which deter-mine minimum and the maximum left ventricular volumes. For R-wave-based ECG-synchronization, end-systole and end-diastole have to be estimated with appropriate phase parameters relative to the R-wave. More advanced ECG-gating techniques that determine the location of the end-systolic phase based on the detection of the T-wave and the end-diastolic phase based on the detection of the P-wave are under investigation. However, given the small signal amplitudes, reliable detection of the T-waves and P-waves requires advanced software-based algorithms, which themselves are currently under development. Nonetheless, the first promis-ing results were obtained in a clinical study that obtained better results with cardiac image recon-struction in end-diastole based on P-wave gating, compared to reconstruction in diastole with a rel-ative delay in relation to the R-wave (Sato 2003) (Fig. 4.38).

It should be kept in mind that the ECG is only an indirect measure of cardiac motion. In some cases, such as in patients with atrial fibrillation or with extra-systoles, the electrical stimulation does not

directly correspond to the real motion of the heart. Therefore, motion-detection algorithms have been investigated that have the potential to directly mea-sure cardiac motion from the acquired scan data.

The first cardiac-motion detection algorithm to be clinically investigated (Ohnesorge 1999) used the opposite parallel-beam projections Pp(Θ,p) and Pp(Θ−π,p) in a fixed image plane measured at a pro-jection angle with a 180° shift. In a static object, the signal difference between two opposite parallel-beam projections measured in the same plane equals zero. If the object is moving, the magnitude of the signal difference can be used to measure the displacement of a particular anatomical structure and the amount of movement that occurs between measurements of the two opposite projections, i.e., within the time of a half rotation of the system (Fig. 4.39). For exam-ple, the sum of the absolute values of the individual signal differences of all complementary rays in the considered opposite projections Σp|Pp(Θ,p)–Pp(Θ–π,p)| can be used as an indicator for the displace-ment and motion of an anatomical structure at the point in time when the projection Pp(Θ,p) was mea-sured. With this approach, it could be demonstrated that the derived signal difference correlates with both the displacement and the motion of the heart during the cardiac cycle (Fig. 4.39). However, the accuracy and consistency of the signal for reproduc-ible detection of equivalent phases in consecutive cardiac cycle is limited, so that use of the algorithm is restricted to assessing the presence or absence of cardiac motion in patients with irregular heart beats during the scan, and therefore as input information for interactive editing of the recorded ECG.

A more recent approach, the so-called Kymogram algorithm (Kachelrieß 2002), derives cardiac motion from projection data based on the calcula-tion of the “center of mass” of the scanned fan-beam projections (mCOM) and the center of mass of the object in the considered image plane. The center of mass mCOM,q of a fan-beam projection Pf(αq,βm) is determined by the ray at angle βm within the projec-tion such that the sum of the signals on both sides of that ray is equivalent (Eq. 4.10). The center of mass of the object in the considered image plane can be determined by the point of intersection of the rays that determine the centers of mass in two closely adjacent projections (Fig. 4.40).

111

50%–RR Trot/2 Trot/2R R R R

T T T

PP

T

Fig. 4.38a, b. Cardiac image quality obtained by a 4-slice CT scanner with 0.5-s rotation time in end-diastole based on R-wave- and P-wave-gated image reconstruction. a During P-wave gating, the end of the image reconstruction window is positioned on the P-wave. b During R-wave gating, the start of the image reconstruction window is positioned at 50% of the RR-interval. P-wave-gated reconstruction yields superior image quality compared to R-wave-gated reconstruction due to more consistent phase selection

a b

waves in ECGs recorded during the scans of patients with regular heart rates, and that Kymogram-gated cardiac image reconstruction could yield results comparable to those obtained with ECG-gated image reconstruction (Fig. 4.42a). However, further clinical studies have demonstrated that Kymogram-gated image reconstruction results in inferior image quality compared to ECG-gated image reconstruc-tion under clinical conditions, in both patients with regular heart rates and in those with irregular heart rates (Fig. 4.42b), (Fischbach 2004).

It can therefore be concluded that ECG-corre-lated scanning and image reconstruction remains

P Pf q mm

m m

f q mm m

m MCOM q

COM q

( , ) ( , );

;

α β α β=

= −

= +

= −

∑ ∑=0

1

1

1

(4.10)

The center of mass changes its position with changing static anatomy from image plane to image plane and with changing positions of an anatomi-cal structure due to motion. After subtracting the function of the center of mass for different slice positions, the variation of the center of mass over time represents a measure of cardiac motion. The resulting function of the variation of the center of mass over time is called the Kymogram (Fig. 4.41). Kachelrieß (2002) demonstrated that peaks of the Kymogram often correlate to the position of R-



Fig. 4.40. Center of mass detection a within a pro-jection and b for the ob-ject from two projections as the basis of Kymogram-gated reconstruction

ECG

MotionCurve

Pp(Θ,p)

|Pp(Θ,p)– Pp(Θ–π,p)|

Pp(Θ–π,p)

Fig. 4.39a, b. The principle of detection of cardiac motion from scan data. a The difference signal of opposite parallel projections is calculated and used as a measure for the motion and displacement of the cardiac anatomy in between measurement of the two projections. A difference signal is determined for every projection as the sum of the absolute values of the individual signal differences of all complementary rays in the considered opposite projections. b The difference signal correlated to the ECG that has been recorded during the scan is shown. Im-ages that are reconstructed during the detected phase of low motion demonstrate fewer motion artifacts

a

b

113

R R

K K

ECG

Kymogram

yc(z)

xc(z)

Fig. 4.41a, b. Principle of the separation of the Kymogram motion function and ana-tomical bias from the center of mass func-tion. a The anatomical bias xc(z), yc(z) of the center of mass within the image plane has to be subtracted from the function of center of mass that depends on z-position and time. b The Kymogram results after subtraction of the anatomical bias from the center of mass function and correlates to the position of the R-waves

b

a

the method of choice to synchronize CT scanning with cardiac motion. If accurate automated P- and T-wave detection algorithms become available, ECG gating related to P- and T-waves will offer a promis-ing alternative to R-wave-related ECG-gating. Other approaches to detect cardiac motion automatically from the projection data of the CT scan have demon-strated the feasibility of the principle, but they have yet to provide sufficient robustness in clinical rou-tine. To date, the use of such algorithms is restricted to a back-up role for the ECG signal, e.g., in case of poor signal detection, R-wave misregistration, or failure of the ECG.

4.6 Radiation Exposure Considerations

Radiation exposure of patients during computed tomography and the resulting potential radiation hazards have recently gained increasing attention,

both in the public and in the scientific literature (Brenner 2001, Nickoloff 2001). In particular, the radiation dose for ECG-synchronized cardiac scan-ning has been a topic of considerable controversy. This section will introduce the basic principles of radiation-dose measurement in CT, exposure esti-mations for ECG-gated cardiac examinations with 16-slice and 64-slice CT systems, and the most recent approaches to reduce radiation exposure during car-diac CT examinations.

4.6.1 Principles of Radiation Dose Measurement in CT

In CT, the average dose in the scan plane is best described by the weighted computerized tomo-graphic dose index (CTDIw) (Morin 2003, McCollough 2003), which is determined from CTDI100 measurements both in the center and at the periphery of a 16-cm Lucite phantom for the head and a 32-cm Lucite phantom for the body. For the CTDI100 measurements, a 100-mm-long ionization chamber

4.6 · Radiation Exposure Considerations


R R R R

Fig. 4.42a, b. Comparison of cardiac image quality of a 16-slice CT scanner with 0.42-s rotation time in a patient with regular sinus rhythm. Reconstruction with a ECG-gating (at 50% of the RR-interval) and b Kymogram-gating (at 50% of the interval between the peaks). The peaks of the Kymogram correlate well with the detected R-Waves. The Kymogram-gated reconstruction demonstrates image quality comparable to the ECG-gated reconstruction except for moderate registration artifacts at the beginning and end of the scan. (Case courtesy of Münster University, Germany)

a b

is used. Figure 4.43 shows the typical equipment for dose measurements. CTDIw, is a good estimate for the average patient dose as long as the patient’s size is similar to that of the respective phantoms. CTDIw, is defined according to (Eq. 4.11) (Morin 2003).

CTDIw = 1/3 CTDI100 (center) + 2/3 CTDI100 (periphery) (4.11)

CTDIw, given in mGy, is always measured in an axial scan mode. It depends on scanner geometry, slice collimation, beam pre-filtration, X-ray tube voltage (in kV), tube current (in mA), and gantry rotation time Trot. The product of mA and Trot is the mAs value of the scan. To obtain a parameter characteristic for the scanner used, it is helpful to eliminate mAs dependence and to introduce a nor-malized (CTDIw)n, given in mGy/mAs:

CTDIw = mA · Trot · (CTDIw)n = mAs · (CTDIw)n (4.12)

CTDIw is a measure of the dose in a single axial scan and depends on X-ray tube voltage and slice collimation. The latter parameters are needed to specify CTDIw. For multi-slice CT systems, CTDIw tends to increase with decreasing collimated slice width, as a consequence of the increasing relative contribution of the penumbra zones of the dose pro-files. As a representative example, Figure 4.44 shows CTDIw at 120 kV for the 32-cm body phantom as a function of the total collimated width of the detector for a 4-slice CT system and a 16-slice CT system with similar system geometry. Scan protocols for differ-ent CT scanners should always be compared on the basis of CTDIw and never on the basis of mAs, since different system geometries can result in significant differences in the radiation dose applied at identical mAs.

(CTDIw)n can be used to calculate the radiation dose for axial scans, both for standard applications and for ECG-triggered sequential scanning. For most CT scanners, the mAs value for axial scans, which is

115

R R R R

Fig. 4.43a, b. Comparison of cardiac image quality of a 16-slice CT scanner with 0.42-s rotation time in a patient with regular sinus rhythm. Reconstruction with a ECG-gating (at 50% of the RR-interval) and b Kymogram-gating(at 50% of the interval between the peaks). The peaks of the Kymogram cor-relate well with some R-waves during the scan but do not correlate well with the detected R-waves throughout the scan. The Kymogram-gated reconstruction demonstrates severe misregistration ar-tifacts while the ECG-gated reconstruction yields artifact-free results. (Case courtesy of Münster University, Germany)

a b

the product of tube current mA and slice exposure time Texp, is indicated on the user interface. Usually, Texp equals the gantry rotation time Trot. For cardiac applications, however, partial scans are generally used for ECG-triggered sequential scanning. In this case, the slice exposure time is about two-thirds of the gantry rotation time. The principle of radiation exposure calculation for ECG-triggered scan proto-cols can be demonstrated with a practical example for ECG-triggered scanning of coronary calcium quantification. The slice exposure time of a 16-slice CT scanner for a partial scan at 0.42-s gantry rota-tion time is about 0.3 s. The scan protocol recom-mended by the manufacturer for ECG-triggered sequential scanning for coronary calcium quanti-fication uses a tube current of 100 mA at a voltage of 120 kV. With these parameters, 100 × 0.3 mAs = 30 mAs are applied. This value is shown on the user interface. (CTDIw)n = 0.072 mGy/mAs can be found in the data sheet of this scanner for the given protocol at 120 kV and for 12 × 1.5-mm collimation

that provides 6 × 3-mm slices per scan. A CTDIw = 0.072 mGy/mAs × 30 mAs = 2.16 mGy can therefore be determined for this protocol (Flohr 2003).

For ECG-gated multi-slice spiral scans of the heart, the situation is more complicated. To represent the dose in a multi-slice spiral scan, it is essential to account for gaps or overlaps between the radiation-dose profiles from consecutive rotations of the X-ray source (Morin 2003). For this purpose, CTDIvol, the volume CTDIw, has been introduced

CTDIvol = 1/pitch · CTDIw = mAs · 1/pitch · (CTDIw)n = mA · Trot · 1/pitch · (CTDIw)n (4.13)

The factor 1/pitch accounts for the increasing dose accumulation with decreasing spiral pitch due to the increasing spiral overlap. In principle, Eq. 4.13 is valid for both single-slice and multi-slice spiral CT. For ECG-gated cardiac scanning, low pitch values (typically pitch = 0.2–0.35, depending on the number of detector rows, the gantry rotation time,



and the number of segments used for image recon-struction) have to be used to ensure gap-less cover-age of the heart volume in all phases of the cardiac cycle. The highly overlapping data acquisition has to be taken into account when calculating the dose for an ECG-gated spiral scan. Again, the principle of radiation exposure calculation for ECG-gated spiral scan protocols can be demonstrated with a practi-cal example of an ECG-gated 16-slice spiral CT scan protocol used for coronary calcium quantification. The example protocol uses a 0.42-s rotation time, 16×1.5-mm collimation, pitch 0.28, 120 kV, and 100-mA tube current. The mAs value for this scan is 100 mA × 0.42 s = 42 mAs, which, in terms of the applied radiation dose, cannot directly be compared to mAs values for ECG-triggered sequential scan-ning, since in the latter the pitch dependency would be neglected. In fact, using CTDIw = 0.070 mGy/mAs, the radiation dose in this case is 0.070 mGy/mAs × 42 mAs × 1/0.28 = 10.5 mGy. For spiral scanning, some manufacturers, such as Siemens and Philips, have introduced the concept of an “effective” mAs, which includes the factor 1/pitch into the mAs defi-nition:

(mAs)eff = mA · Trot · 1/pitch = mAs · 1/ pitch (4.14)

For spiral scans, (mAs)eff is indicated on the user interface. Inserting Eq. 4.14 into Eq. 4.13, the dose of a multi-slice spiral CT scan is given by

CTDIvol = (mAs)eff x (CTDIw)n (4.15)

Other manufacturers (i.e., Toshiba and GE) have retained the conventional mAs definition, so that the user has to perform the 1/pitch correction. When comparing the scan parameters for CT systems of different manufacturers, the underlying mAs defi-nition has to be taken into account. In the above example of an ECG-gated multi-slice spiral scan for coronary calcium quantification, the user applies 42 mAs, but 42 mAs × 1/0.28 = 150 effective mAs.

CTDIw is a measure of physical dose; it does not pro-vide full information on the radiation risk associated with a CT examination. For this purpose, the concept of “effective dose” has been introduced by the ICRP (International Commission on Radiation Protection). The effective dose is given in milli-Sieverts (mSv). It is a weighted sum of the dose applied to all organs in a CT examination and includes both direct and scattered radiation. The weighting factors depend on the biological radiation sensitivities of the respective organs. Effective dose can be measured using whole-body phantoms, such as the Alderson phantom, or it is obtained by computer simulations using Monte Carlo techniques to determine scattered radiation. The effective patient dose depends on the scanned range. For a comparison of effective dose values for different protocols, scan ranges should be similar. A suitable program for calculation of effective dose values is, e. g., WinDose (Kalender 1999). This is a PC-based program that calculates organ dose and effective dose values for arbitrary scan parameters and anatomical ranges. Values for primary radiation have to be measured in terms of CTDIw and are used as input; values for scattered radiation are derived from Monte Carlo calculations.

The user has to keep in mind that both CTDIw and effective patient dose, which is derived from CTDIw, give a correct estimation of the radiation dose for the patient only when the cross-section of the patient’s anatomy is comparable in size to the cross-section of the phantom used for the evaluation of CTDIw (32-cm Lucite for the body). For smaller patients, the dose may be considerably higher (Fig. 4.45).

Fig. 4.44. Dose measurements in CT. The equipment consists of a 100-mm-long ionization chamber and a 32-cm Lucite phantom. With this set-up, the CTDIw value for the body is determined

117

4.6.2 Radiation Exposure for Selected Cardiac Examination Protocols

Typical values for the effective patient dose of selected multi-slice CT protocols are 1–2 mSv

for a head examination, 5–7 mSv for a chest CT, and 8–11 mSv for CT of the abdomen and pelvis (McCollough 2003, Morin 2003). This radiation exposure must be appreciated in the context of the average annual background radiation, which is 2–5 mSv (about 3.6 mSv in the US).

This section provides estimates of the effective patient dose for selected cardiac CT protocols with 16-slice and 64-slice CT systems, using the Siemens SOMATOM Sensation 16 and Sensation 64 scanners as representative examples. The effective patient dose for the standard protocols recommended by the manufacturer was calculated with WinDose (Kalender 1999). The scan protocols included are ECG-triggered coronary calcium quantifica-tion, ECG-gated coronary calcium quantification, and ECG-gated coronary CT angiography. The scan range used is 12 cm for coronary calcium quantifi-cation and 10 cm for coronary CT angiography. The scan parameters recommended by the manufacturer are listed in Table 4.2. For sequential scans, the mAs values are based on the slice exposure times for an axial scan that is the partial-scan time. For spiral

10CTDIW 4-Slice CT versus 16-Slice CT

CTD

I W in

mG

y /

100m

As 9

8

7

6

5

40 4 8 12 16 20 24 28

2x1mm4x1mm

16x0.75mm

4x2.5mm

4x5mm

16x1.5mm

Total collimation in mm

Fig. 4.45. CTDIw at 120 kV for the 32-cm body phantom as a function of the total collimated width of the detector for 4-slice (black line) and 16-slice (blue line) CT systems with similar geometries


Table 4.2. Scan protocols used for calculation of the effective patient dose

ECG-triggered calcium scoring

ECG-gated calcium scoring ECG-gated cardiac/coronary CT angiography

Sensation 16 Sensation 64 Sensation 16 Sensation 64 Sensation 16 Sensation 64

Collimation (mm)

12 × 1.5 30 × 0.6 16 × 1.5 24 × 1.2 16 × 0.75 32 × 0.6*

(*64 × 0.6 by z-FFS)

Recon-structed slice (mm)

3.0 3.0 3.0 3.0 0.75/1.0 0.6/0.75

Rotation time (s)

0.42 0.375 0.375 0.33 0.42 0.375 0.375 0.33 0.42 0.375 0.375 0.33

kV 120 120 120 120 120 120 120 120 120 120 120 120

mA 100 111 111 126 100 113 109 115 370 413 435 467

mAs 30 30 30 30 42 42 41 38 155 155 163 154

Pitch n.a. n.a. 0.28 0.25 0.24 0.20 0.28 0.25 0.24 0.20

Table feed 18 mm per axial scan

18 mm per axial scan

16 mm/s 16 mm/s 13.2 mm/s 16 mm/s 8 mm/s 8 mm/s 12.3 mm/s 11.7 mm/s

Effective mAs n.a. n.a. n.a. n.a. 150 170 170 190 555 620 680 770

(CTDIw)n (mGy/mAs)

0.072 0.072 0.072 0.072 0.070 0.070 0.067 0.067 0.078 0.078 0.077 0.077

CTDIw (mGy) 2.2 2.2 2.2 2.2 10.5 11.9 11.4 12.7 43.3 48.4 52.4 59.3


scans, effective mAs values are used; these have been already corrected for spiral pitch, i. e., they take into account the dose accumulation with decreas-ing spiral pitch (see above). Table 4.3 summarizes the CTDIw values for the 32-cm body-phantom and values of the effective patient dose for the recom-mended standard protocols, for the 16- and the 64-slice CT systems. In Table 4.3, the scan parameters listed in Table 4.2 were used.

For ECG-triggered coronary calcium quanti-fication with sequential scanning, patient dose is about 0.5 mSv for males and 0.68 mSv for females. Coronary calcium quantification with ECG-gated spiral scanning that results in reduced interscan variability (Ohnesorge 2002) increases the patient dose to 2.3–2.8 mSv for males and 3.3–3.9 mSv for females. ECG-gated high-resolution coronary CT angiography using thin-slice spiral data acquisi-tion and yielding both adequate visualization of the small and complex cardiac and coronary anatomy and detection and classification of coronary plaques requires a dose of 7.9–10.8 mSv for males and 11.1–15.2 mSv for females. Improved longitudinal resolu-tion by thinner collimation directly translates into

increased patient dose if the signal-to-noise ratio of the images is maintained.

A comparison of radiation doses for different CT scanner types requires that imaging parameters, such as slice width, tube voltage, tube current, and scanned volume, be taken into account. If the high spatial resolution of modern multi-slice CT systems achieved with sub-millimeter slice collimation is not required, decreased longitudinal spatial resolu-tion with slice-widths between 1.5 and 3.0 mm can always be traded off with a correspondingly reduced patient dose.

The pitch values in the ECG-gated multi-slice spiral CT protocols shown in Table 4.2 are selected such that the vast majority of patients can be scanned without significant protocol adjustments, including patients with low heart rates. Radiation exposure, however, can be reduced by increasing the pitch in patients with sufficiently high heart rates when respecting the constraint considerations given in Section 4.4.3 and in Figure 4.16. The radia-tion exposure is then reversed proportional to the pitch increase, as compared to the values presented in Tables 4.2 and 4.3.

Table 4.3. Values for the effective patient dose based on the scan protocols listed in Table 4.2

ECG-triggered calcium scoring ECG-gated calcium scoring ECG-gated cardiac/coronary CT angiography

Sensation16 Sensation64 Sensation16 Sensation64 Sensation16 Sensation64

Collimation ( mm) 12 × 1.5 30 × 0.6 16 × 1.5 24 × 1.2 16 × 0.75 32 × 0.6 *

(*64 × 0.6 by z-FFS)

Rotation time ( s) 0.42 0.375 0.375 0.33 0.42 0.375 0.375 0.33 0.42 0.375 0.375 0.33

CTDIw ( mGy) 2.2 2.2 2.2 2.2 10.5 11.9 11.4 12.7 43.3 48.4 52.4 59.3

Scan range ( mm) 120 120 120 120 100 100

Effective patient dose ( mSv), males without ECG-pulsing

0.5 0.5 0.5 0.5 2.3 2.6 2.5 2.8 7.9 8.9 9.6 10.8

Effective patient dose ( mSv), females without ECG-pulsing

0.68 0.68 0.68 0.68 3.2 3.7 3.5 3.9 11.1 12.4 13.5 15.2

Effective patient dose ( mSv), males with ECG-pulsing

n.a. n.a. n.a. n.a. 1.2–1.6

1.3–1.8

1.3–1.8

1.4–2.0

4.0–5.5

4.5–6.2

4.8–6.7 5.4–7.6

Effective patient dose ( mSv), female with ECG-pulsing

n.a. n.a. n.a. n.a. 1.6–2.2

1.9–2.6

1.8–2.5

2.0–2.7

5.6–7.8

6.2–8.7

6.8–9.4 7.6–10.6

119

4.6.3 Exposure Reduction with ECG-Gated Tube-Current Modulation

The relatively high radiation exposure for ECG-gated multi-slice spiral imaging of the heart is caused by continuous X-ray exposure and data acquisition at low and highly overlapping spiral pitch (pitch << number of slices). The low spiral pitch is a conse-quence of the phase-consistent coverage of the heart volume in specific phases of the cardiac cycle. How-ever, if the spiral pitch is limited such that one phase can be covered consistently, all other phases of the cardiac cycle can be covered as well. If reconstruc-tion in different cardiac phases is not needed and instead only a very limited interval (i.e., diastolic phase) in the cardiac cycle is targeted during recon-struction, a significant portion of the acquired data and radiation exposure is redundant.

Prospective ECG triggering combined with “step-and-shoot” acquisition of axial slices has the ben-efit of smaller patient dose compared to ECG-gated spiral scanning, since scan data are acquired in the previously selected heart phases only. It does, how-ever, not provide continuous volume coverage with overlapping slices and misregistration of anatomi-cal details cannot be avoided. Furthermore, recon-struction of images in different phases of the cardiac cycle for functional evaluation is not possible. Since ECG-triggered axial scanning depends on a reliable prediction of the patient’s next RR-interval based on the mean of the preceding RR-intervals, the method encounters its limitations for patients with severe arrhythmia. To maintain the benefits of ECG-gated spiral CT but reduce patient dose ECG-controlled dose modulation has been developed. On-line reduc-tion of the tube output in each cardiac cycle during phases that are of less importance for ECG-gated reconstruction has a high potential for exposure reduction. The nominal tube output is only required during those phases of the cardiac cycle that will be reconstructed. During ECG-gated tube-current modulation, the tube output is modulated on-line with prospective ECG control (Ohnesorge 2000b, Jacobs 2002, Poll 2002), and commonly used ECG-gated reconstruction algorithms can be further used unchanged. Within every cardiac cycle, tube output is raised to the nominal level during a limited inter-

val in a pre-selected phase (usually the diastolic phase) in which the data are most likely to be recon-structed with thin slices and a high signal-to-noise ratio needs to be maintained. During the remain-ing part of the cardiac cycle, the tube output can be reduced by about 80% by a corresponding decrease of the tube current mAmin = 0.2 mAmax (Fig. 4.46). Thus, continuous volume reconstruction is still possible in all phases of the cardiac cycle. In par-ticular, functional imaging is still feasible, as it does not require thin slice reconstruction. For imaging during phases of reduced tube output, an appropri-ate signal-to-noise ratio can be maintained by pri-mary or secondary reconstruction of thicker slices. The position of the windows of nominal tube output within the heart cycles needs to be defined prior to the scan in relation to the phase targeted for recon-struction.

The width of the time interval ∆TN with nominal tube output during diastole has to be selected such that patient-individual shifting of the ECG gating interval is still possible to obtain the best possible image quality. Additionally, a well-defined over-lap of ∆TN with the window of temporal resolution (i.e., 165–250 ms) can compensate for inconsistent prospectively ECG-controlled timing of nominal tube output due to changes in heart rate during the scan. In addition, an extended length of the temporal window with nominal tube output allows for high-resolution reconstruction within an extended dura-tion during the cardiac cycle; this may be useful for optimization of image quality. An appropriate trade-off of exposure reduction and ECG gating flexibility can be achieved with the selection of ∆TN = 400 ms. Shorter ∆TN = 300 ms for further exposure reduction may be possible for scanners with very fast rotation times (≥0.37 s). However, the possibility to recon-struct high-resolution images within an extended range of the cardiac cycle is then limited. The relative exposure reduction with ECG-controlled tube-cur-rent modulation as a function of heart rate is shown in Figure 4.47. For normal heart rates between 50 and 90 bpm, the exposure is reduced by 35–50% for ∆TN = 400 ms and by 45–60% for ∆TN = 300 ms. Radiation exposure savings are maximized for low heart rates, as the total time with low tube output during the scan is high. For increasing heart rates, the relative reduc-tion decreases, as the time intervals of low tube output



Reco

n

Reco

n

Time

z - P

osit

ion Detector

Width

Reco

n

R R R R

Tdel,mA

Tdel,Recon

∆TN

Continuous

Spiral Scan & FeedLast Detector Row

First Detector Row

“3D”Image DataTrot / 2 Temp. Res.

100%

20%

mA

Fig. 4.46. Retrospectively ECG-gated 4-slice spiral scanning with prospectively ECG-controlled tube-current modulation for reduced radiation exposure. Tube current is at a nominal value (time interval ∆TN) during diastole and is reduced by 80% during phases of high cardiac motion

Exp

osu

re R

edu

ctio

n [%

]

120

Heart Rates 50–90bpm

Heart Rate in BPM110100908070605040

80

70

60

50

40

30

20

10

0

∆TN = 300msmAmin = 0.2mAmax

∆TN = 400msmAmin = 0.2mAmax

Fig. 4.47. Exposure reduction with ECG-controlled tube-cur-rent modulation dependent on heart rate for different time intervals ∆TN of nominal tube output. A 35–55% reduction of exposure can be achieved for regular heart rates between 50 and 90 bpm using ∆TN = 400 ms

are shorter. However, for very high heart rates, the selection of a higher spiral pitch can decrease expo-sure via shorter scan time as compared to low heart rates. Figure 4.48 directly compares two consecutive scans for quantification of coronary calcification in the same patient. The scans were performed without and with ECG-controlled tube output modulation, and equivalent reconstructions were carried out in the diastolic phase. For the patient with a stable heart rate between 58 and 61 bpm, an exposure reduction of 51% could be achieved with the same image and diagnostic qualities of the clinical result. An example for a CT angiography scan of the heart and coronary arteries using ECG-controlled tube-current modu-lation is shown in Figure 4.49. High signal-to-noise ratio is available for reconstruction during diastole with thin sections and high resolution. Reconstruc-tion during systole for functional information shows lower signal-to-noise ratio; however, diagnostically sufficient image quality can be achieved with sec-ondary multi-planar reformation (MPR) recon-struction. For the patient with a stable heart rate

121

58bpm 59bpm 60bpm

20mA 20mA 20mA

100mA 100mA 100mA

Fig. 4.48a–c. Case example of two repeated ECG-gated spiral scans of the same patient using a 4-slice CT scanner at 0.5-s rotation time and 4 × 2.5-mm collimation for imaging calcifi ed plaques in the coronary arteries by applying ECG-controlled tube-current modulation. The patient showed stable sinus rhythm at heart rate 59 bpm during both scans. Calcifi cations in the left anterior descending and right coronary arteries that were identifi ed with the fi rst scan (a) can be accurately reproduced with the second scan (b). By ECG-controlled modulation of the tube current between 100 mA during diastole and 20 mA during systole (c), the amount of radiation can be reduced by 51% without compromising the signal-to-noise ratio for 59 bpm average heart rate. (Images courtesy of Klinikum Grosshadern, Munich, Germany)

a b

c

between 74 and 77 bpm, the exposure reduction with ECG-gated tube-current modulation was about 37%. Clinical studies based on patients separated in two groups with similar patient parameters (i.e., patient size, heart rate variations, median heart rates) have demonstrated dose reductions of 40–50% in male and female patients without remarkable influence on image noise and image quality by using ECG-con-trolled tube-current modulation (Jakobs 2002).

Where available, this technique should be used for all patients with reasonably steady heart rates (Trabold 2003). However, if substantial arrhyth-mia is expected during the scan, prospective control of the tube-current modulation and retrospective positioning of the reconstruction intervals may not match and ECG-controlled tube-current modula-

tion may not be adequately applied. To allow for more reliable use of ECG-controlled tube-current modulation in patients with irregular heart rates, new developments are under way that employ intel-ligent ECG analysis algorithms that detect extra-systolic beats in real time and disenable or adjust the temporal window for tube-current reduction in the affected heart beats (Fig. 4.50).

4.6.4 Optimization for Different Patient Sizes

In both ECG-triggered sequential scanning and ECG-gated spiral scanning, adaptation of the dose to patient size and weight is an important potential for



76bpm 75bpm 75bpm

94mA 94mA 94mA

467mA 467mA 467mA

+20% +20% +20%+70% +70% +70%

End-Diastole +70% End-Systole +20%

Fig. 4.49a–c. Example of a coronary CT angiography examination using a 64-slice CT scanner with 64 × 0.6-mm slices, 0.33-s rotation time, and ECG-controlled dose modulation. a End-diastolic image with full dose, reconstructed at 70% of the cardiac cycle; b end-systolic image with reduced dose, reconstructed at 20% of the cardiac cycle; c The ECG signal of the patient shows a stable heart rate between 74 and 77 bpm. Although the signal-to-noise ratio of the end-systolic image is reduced, it is considered adequate for functional evaluation. (Images courtesy of University of Erlangen, Germany)

a b

c

75bpm

20%

+60% +60% +60% +60%

+60% +60% +60% +60%

20% 20% 20%

100% 100% 100% 100%

105bpm 45bpm 75bpm

75bpm 105bpm 45bpm 75bpm

20% 20%

100% 100%

Fig. 4.50a, b. Automatic detection of extra-systolic heart beats during ECG-controlled tube-current modulation in a patient with a regular heart rate of 75 bpm. a An extra-systolic heart beat may lead to a mismatch in several consecutive heart beats of the temporal window with maximum tube output and desired cardiac phase for high-resolution image reconstruction. b If an extra-systolic heart beat is detected, the reduction of tube current can be disenabled during the affected heart beats and image quality can be maintained throughout the entire scan

a

b

123

dose reduction. For quantification of coronary cal-cifications, a reduction of the mean effective patient dose by 11.8% for males and 24.8% for females in a group of 50 patients was demonstrated by indi-vidual body-weight-adapted tube-current settings (Mahnken 2003). For cardiac and coronary CT angiography, a similar potential for radiation dose reduction has been reported (Jung 2003). The Inter-national Consortium of Standardization in Cardiac CT, a group of scientists, physicians, and CT manu-facturers, has published recommendations to gener-ate more accurate and calibrated results for coronary artery calcification independent of the scanner and of patient size. The researchers used an anthropo-morphic calibration phantom with fixed amounts of calcium to derive dose recommendations for small, medium-sized, and large patients. The anthropo-morphic calibration phantom consisted of tissue-, water- and bone-equivalent materials (QRM, Qual-ity Assurance in Radiology and Medicine GmbH, Möhrendorf, Germany) and is shown in Figure 4.51 (Ulzheimer 2003). To account for different patient sizes, the original phantom can be used with up to two additional attenuation rings. The three catego-ries for patient size are defined as small: < 32.0-cm lateral thickness, medium: 32.0- to 38.0-cm lateral thickness, and large: > 38.0-cm lateral thickness. An A/P localizer image (topogram, scout view) is used to determine the lateral thickness, which is measured from skin-to-skin, at the level of the proximal ascending aorta. Target noise levels of 20 HU for small and medium-sized patients and 23 HU for large patients in the corresponding calibra-tion phantom are proposed. For ECG-gated spiral scanning of coronary artery calcification using a 16-slice CT scanner with 16 × 1.5-mm collima-tion, 0.42-s rotation time, and 120 kV, the follow-ing effective mAs values are feasible: small = 80–85 effective mAs, medium = 200–240 effective mAs, and large = 450–500 effective mAs. Similar rela-tions can be found for high-resolution and cardiac coronary CT angiography protocols in which target noise levels of 15–20 HU with thin-slice reconstruc-tion of 1 mm and below need to be achieved. The above example demonstrates the high potential for dose reduction, especially in medium-size and small patients, for coronary calcium quantification as well as cardiac and coronary CTA imaging protocols. In

some modern CT scanners, the mAs value applied during an ECG-gated spiral scan can be automati-cally adapted to the average size of the patient (e.g., CARE Dose4D, Siemens, Forchheim, Germany).

4.6.5 Optimization of Contrast-to-Noise Ratio

Another means to reduce the radiation dose is to adapt the X-ray tube voltage to the intended appli-cation. Typically, 120 kV are used for quantifica-tion of coronary calcifications and for coronary CT angiography. In these two applications, however, the contrast-to-noise ratio for fixed patient dose can be increased by decreasing the X-ray tube voltage. As a consequence, to obtain a desired contrast-to-noise ratio, the patient dose can potentially be reduced by choosing lower kV settings of 80 or 100 kV. The potential for dose saving is more pronounced for small and medium-sized patients. Phantom mea-surements using small tubes filled with diluted contrast agent embedded in Lucite phantoms with different diameters (Schaller 2001) were used to demonstrate the potential to increase contrast and reduce dose in general vascular and cardiac CT angi-ography examinations in small and medium-sized patients when using 100 kV. Initial clinical experi-ences have confirmed this finding in patients weigh-ing up to 90 kg (Fig. 4.52). Other studies (Jacobs 2003) compared ECG-gated spiral scan protocols using 120 and 80 kV for the quantification of cor-onary calcium. An 80-kV scan protocol was used in combination with ECG-controlled tube-current modulation, resulting in a significantly lower patient radiation exposure (0.72 mSv for 80 kV compared to 2.04 mSv for 120 kV). CTDI phantom measurements revealed a 65% reduction of radiation dose with the 80-kV protocol. In this study, the 80-kV protocol led to a significant reduction in radiation expo-sure during multi-slice CT coronary artery calcium screening, but did not affect the detection and quan-tification of coronary artery calcification. However, it should be noted that the maximum X-ray tube current available at 80 kV is generally not sufficient to scan larger patients and an increase in the tube voltage is required. In such patients, use of the 100-kV tube voltage may be a good compromise.



Fig. 4.52. Case example of a CT angiographic ex-amination of the coronary arteries in a patient weighing 82 kg . The image was obtained on a 64-slice CT scanner with 64 × 0.6-mm slice acquisi-tion and a 0.33-s rotation time. To reduce radia-tion exposure, 100-kV tube voltage and ECG-con-trolled tube-current modulation were used. The use of 100 kV instead of 120 kV reduces radiation exposure by about 33%, from 6.8 to 5.0 mSv. The high contrast-to-noise ratio allows visualization of the entire 3D cardiac and coronary anatomy and of atherosclerotic plaques in the left anterior descending and circumfl ex coronary arteries us-ing an overlay of VRT and MIP displays. (Images courtesy of Medical University of South Carolina, Charleston, USA)

Fig. 4.51. Anthropomorphic calibration phantom (QRM Qual-ity Assurance in Radiology and Medicine GmbH, Möhrendorf, Germany) recommended by the International Consortium for Standardization in Cardiac CT. The original phantom can be used with up to two additional attenuation rings to simulate small, medium-sized, and large patients

References

Achenbach S, Ropers D, Holle J, Muschiol G, Daniel WG, Moshage W (2000a). In-plane coronary arterial motion velocity: measurement with electron-beam CT. Radiology 216:457–463.

Bahner ML, Böse J, Lutz A, Wallschläger H, Regn J, van Kaick G. (1999). Retrospectively ECG-gated spiral CT of the heart and lung. Eur Radiol 9:106–109

Becker CR, Knez A, Jakobs TF, Becker A, Schöpf UJ, Brüning R, Haberl R, Reiser MF (1999). Detection and quantification of coronary artery calcification with electron-beam and conventional CT. Eur Radiol 9:620–624

Brenner D, Elliston C, Hall E, Berdon W (2001). Estimated risks of radiation-induced fatal cancer from pediatric CT. AJR 176:289 –296

Bruder H, Stierstorfer K, Ohnesorge B, Schaller S, Flohr T (2001). Segmented cardiac volume reconstruction – a novel reconstruction scheme for multi-slice cardiac CT. The Sixth International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine, Pacific Grove, 2001, 161–164

Bruder H, Stierstorfer K, Ohnesorge B, Schaller S, Flohr T (2002). A novel reconstruction scheme for cardiac volume imaging with MSCT providing cone correction. SPIE Transactions on Medical Imaging

Flohr T, Schaller S, Ohnesorge B, Kopp AF, Klingenbeck-Regn K (1999). Evaluation of image artifacts in multi-slice spiral CT (abstract). Radiology 213(P):317

Flohr T, Ohnesorge B (2001). Heart-rate adaptive optimiza-

125

tion of spatial and temporal resolution for ECG-gated multi-slice spiral CT of the heart. J Comp Assist T 25(6): 907–923

Flohr T, Prokop M, Schöpf, Kopp A, Becker C, Schaller S, White R, Ohnesorge B (2002). A new ECG-gated multi-slice spiral CT scan and reconstruction technique with extended volume coverage for cardio-thoracic applications. Eur Radiol 12:1527–1532

Flohr T, Ohnesorge B, Bruder H, et al (2003). Image recon-struction and performance evaluation for ECG-gated spiral scanning with a 16-slice CT system. Med Phys 30(10):2650–2662

Flohr TG, Schöpf UJ, Küttner A, Halliburton SS, Bruder H, Süß C, Schmidt B, Hofmann L, Yucel EK, Schaller S, Ohnesorge BM (2003). Advances in cardiac multi-slice CT-imaging with 16-slice CT-systems. Acad Radiol 10: 1050–1058

Fischbach R (2004). Clinical evaluation of Kymogram-gated cardiac image reconstruction with 16-slice CT (abstract). Radiology 231(P)

Hong C, Becker,CR, Huber A, Schöpf UJ, Ohnesorge B, Knez A, Brüning R, Reiser MF (2001a). ECG-gated reconstructed multi-detector row CT coronary angiography: effect of varying trigger delay on image quality. Radiology 220:712–717

Hu H (1999). Multi-slice helical CT: scan and reconstruction. Med Phys 26 (1) 5–18

Hu H, Pan T, Shen Y (2000). Multi-slice helical CT: image tem-poral resolution. IEEE Transactions on Medical Imaging vol 19(5):384–390

Jakobs T F, Becker C R, Ohnesorge B, Flohr T, Suess C, Schoepf U J, Reiser MF (2002). Multi-slice helical CT of the heart with retrospective ECG-gating: reduction of radiation exposure by ECG-controlled tube current modulation. Eur Radiol 12: 1081–1086

Jakobs T F, Wintersperger B J, Herzog P, Flohr T, Suess C, Knez A, Reiser MF, Becker C R (2003). Ultra-low-dose coronary artery calcium screening using multi-slice CT with retro-spective ECG-gating. Eur Radiol 13: 1923–1930

Jung B, Mahnken A H, Stargardt A, Simon J, Flohr T G, Schaller S, Koos A, Guenther R W, Wildberger J E (2003). Individually weight-adapted examination protocol in retrospectively ECG-gated MSCT of the heart. Eur Radiol 13: 2560–2566

Kachelrieß M, Kalender WA (1998). Electrocardiogram-cor-related image reconstruction from subsecond spiral computed tomography scans of the heart. Med Phys 25: 2417–2431

Kachelrieß M, Ulzheimer S, Kalender WA (2000). ECG-cor-related image reconstruction from subsecond multi-row spiral CT scans of the heart. Med Phys 27:1881–1902

Kachelrieß M, Sennst DA, Maxlmoser W, Kalender WA (2002). Kymogram detection and kymogram-correlated image reconstruction from subsecond spiral computed tomogra-phy scans of the heart. Med Phys 29(7):1489–1503

Kak AC, Slaney M (1998). Principles of computerized tomo-graphic imaging. IEEE Press New York, p. 77–86

Kalender W (1995). Thin-section three-dimensional spiral CT: is isotropic imaging possible? Radiology 197:578–580

Kalender W A, Schmidt B, Zankl M, Schmidt M (1999). A PC Program for estimating organ dose and effective dose values in computed tomography. Eur Radiol 9:555–562

Klingenbeck K, Schaller S, Flohr T, Ohnesorge B, Kopp AF, Baum U (1999). Subsecond multi-slice computed tomog-raphy: basics and applications. Eur J Radiol 31(2):110–124

Kopp AF, Küttner A, Heuschmid M; Schröder S, Ohnesorge B, Claussen CD (2002). Multidetector-row CT cardiac imag-ing with 4 and 16 slices for coronary CTA and imaging of atherosclerotic plaques. Eur Radiol 12(Suppl.2):S17-S24

Loubeyre P, Angelie E, Grozel F, Abidi H, Minh VA (1997). Spiral CT artifact that simulates aortic dissection: image recon-struction with use of 180 degrees and 360 degrees linear-interpolation algorithms. Radiology 205:153–157

Mahnken A, Jung B, Wildberger J, et al (2003). Patient-size adapted scan protocols in coronary calcium scoring with multi-slice CT. Eu Radiol 13

McCollough C (2003). Patient dose in cardiac computed tomography. Herz 28:1–6

Mochizuki T, Murage K, Higashino H et al (2000). Two and three dimensional CT ventriculography: a new application of helical CT. AJR 174: 203–208

Morin R, Gerber T, McCollough C (2003). Radiation dose in computed tomography of the heart. Circulation 107:917–922

Nickoloff E, Alderson P (2001). Radiation exposure to patients from CT: reality, public perception, and policy. AJR 177:285–287

Ohnesorge B, Flohr T, Rienmüller R (1999). Procedure to deter-mine the reconstruction window in CT examinations of the heart. European Patent Office

Ohnesorge B, Flohr T, Becker C R, Kopp A F, Knez A Baum U, Klingenbeck-Regn K Reiser MF (2000a). Cardiac imaging by means of electrocardiographically gated multisection spiral CT: initial experience. Radiology 217:564–571

Ohnesorge B, Flohr T, Becker CR, Kopp AF, Knez A, Reiser MF (2000b). Dose evaluation and dose reduction strategies for ECG-gated multi-slice spiral CT of the heart (Abstract). Radiology 217(P):487

Ohnesorge B, Flohr T, Becker CR, Knez A, Schöpf U, Klingen-beck K, Brüning R, Reiser MF (2001). Technical Aspects and Applications of Fast Multislice Cardiac CT. Medical Radiology, Diagnostic Imaging and Radiation Oncology, Multislice CT. Springer Verlag Berlin, Heidelberg. Chapter 15:121–130

Parker DL (1982). Optimal short scan convolution reconstruc-tion for fanbeam CT. Med Phys 9(2):254–257

Poll L, Cohnen M, Brachten S, Ewen K, Modder U (2002). Dose reduction in multidetector row CT of the heart by use of ECG-controlled tube current modulation (“ECG-Pulsing”): phantom measurements. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 174:1500–5

Sato Y, Kanmatsuse K, Inoue F, et al (2003). Noninvasive coro-nary artery imaging by multi-slice spiral computed tomog-raphy – a novel approach for a retrospectively ECG-gated reconstruction technique. Circ J 67: 107–111

References


Schaller S, Flohr T, Klingenbeck K, Krause J, Fuchs T, Kalender WA (2000). Spiral interpolation algorithm for multi-slice spiral CT. Part I: Theory. IEEE Transactions on Medical Imaging, 19(9):822–834

Schaller S, Niethammer M U, Chen X, Klotz E, Wildberger J E, Flohr T (2001). Comparison of signal-to-noise and dose values at different tube voltages for protocol optimization in pediatric CT (Abstract). Radiology 223(P), 366

Schöpf UJ, Helmberger T, Holzknecht N, Kang DS, Brüning RD, Aydemir S, Becker CR, Mühling O, Knez A, Haberl R, Reiser MF (2000). Segmental and subsegmental pulmonary arteries: evaluation with electron beam CT versus spiral CT. Radiology 214:433–439

Stanford W, Rumberger J (1992) Ultrafast computed tomog-

raphy in cardiac imaging: principles and practise. Futura Publishing Company New York

Stehling MK, Turner R, Mansfield P (1991). Echo-planar imag-ing: magnetic resonance imaging in a fraction of a second. Science 254: 43–50

Taguchi K, Aradate H (1998). Algorithm for image reconstruc-tion in multi-slice helical CT. Med Phys 25 (4):550–561

Trabold T, Buchgeister M, Küttner A, Heuschmid M, Kopp AF, Schröder S, Claussen CD (2003). Estimation of radiation exposure in 16-detector row computed tomography of the heart with retrospective ECG-gating. RöFo 175:1051–1055

Ulzheimer S, Kalender W A (2003). Assessment of calcium scoring performance in cardiac computed tomography. Eur Radiol 13(3): 484–97

principles of multi-slice cardiac ct imaging 4

Documents